Use LBA scatter lists to calculate the addresses corresponding to the entries in the LBA scatter lists.
By using LBA scatter lists, the I/O operand problem of SSDs under 4K granularity constraints is solved, achieving more efficient data access and interconnect throughput, reducing costs, and making it suitable for deep queues and high-latency environments for graphics processing units.
Patent Information
- Application Number
- CN202480090949.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-07-15
- Filing Date
- 2024-10-29
- Publication Date
- 2026-11-17
AI Technical Summary
In existing technologies, solid-state drives (SSDs) are limited by 4K granularity when accessing data, which restricts the number of I/O operations to non-linear expansion. This makes it impossible to effectively utilize the zero-overhead context switching model of graphics processing units (GPUs), and existing solutions are either costly or inefficient.
By employing Logical Block Addressing (LBA) scatter lists, multiple read requests are aggregated by allocating an LBA scatter list array, an available LBA scatter list index is selected, the LBA list is populated, and commands are sent to the target storage device to calculate the corresponding address, enabling finer-grained data access.
It improves the efficiency of I/O operations and interconnect throughput, reduces costs, enables deeper queues and parallel synchronization, and improves the efficiency of data access.
Smart Images

Figure CN123125568A_ABST
Abstract
Description
priority
[0001] This application claims priority to U.S. Provisional Patent Application No. 63 / 640,525, filed April 30, 2024, the contents of which are incorporated herein by reference in their entirety. Technical Field
[0002] This disclosure relates to accessing data on a non-volatile device, and more specifically to using a Logical Block Addressing (LBA) scatter list to compute an address corresponding to an entry in the LBA scatter list. Background Technology
[0003] Solid-state drives (SSDs) typically use 4096-byte (4K) sector sizes (called blocks) to achieve read and write performance. Central processing unit (CPU)-based systems typically process data in larger blocks. In traditional servers, the available input / output operations per second (IOP / s) are limited by non-linear scaling due to context switching overhead. Graphics processing units (GPUs) can have a zero-overhead context switching model, which can easily consume high throughput of smaller data units. Due to the GPU's grid architecture, similar requests for the same type of data will be handled by a grid of GPU execution threads.
[0004] Therefore, while SSD hardware features front-end non-volatile memory high-speed (NVMe) pipeline logic based on 4K performance targets, implementing lower granularity (such as 512-byte or 256-byte granularity) will use 8 to 16 times the IOP / s target of 4K granularity. Some artificial intelligence (AI) applications use large memory spaces, and implementing such memory spaces using dynamic random access memory (DRAM) is costly. Therefore, currently available solutions are either too slow or use too many drives to concentrate 4K accesses into 256 bytes. Summary of the Invention
[0005] Examples of this disclosure may include an article of manufacture for calculating addresses corresponding to entries in a logical block addressing (LBA) scatter list using a logical block addressing (LBA) scatter list. The article of manufacture may include non-transitory memory including machine-readable instructions that, when executed by a processor, cause the processor to allocate an array of logical block arrays (LBAs). Additionally, the instructions may cause the processor to transmit an initialization command to a target storage device, the initialization command providing the target storage device with addresses pointing to the array of LBAs. The instructions may also cause the processor to aggregate multiple read requests and select an available LBA scatter list index. The instructions may further cause the processor to: populate an LBA list indicated by the selected LBA scatter list index; and transmit subsequent commands to the target storage device to cause the target storage device to calculate an address corresponding to a given entry in the selected LBA scatter list index.
[0006] In conjunction with any of the examples above, allocating the LBA scatter list array may include: setting the size of the LBA scatter list array, the maximum number of entries in the LBA scatter list array, and the number of lists in the LBA scatter list array.
[0007] Combining any of the examples above, this processor can be used to specify the size of the offset.
[0008] Combining any of the examples above, populating the LBA list could include filling the LBA list with the multiple read requests.
[0009] Combining any of the examples above, the given entry in this LBA list could include START_OFFSET and END_OFFSET.
[0010] Combining any of the examples above, the processor can be used to: receive data from the target storage device corresponding to data requested in the given entry of the LBA scatter list index; and release the selected LBA scatter list index.
[0011] Combining any of the examples above, this subsequent command enables the target storage device to: calculate the LBA scatter list address based on the base host address and the LSLI index; and request a DMA transfer from host memory to the LBA scatter list shadow buffer.
[0012] Individually or in combination with any of the examples above, examples of this disclosure may include a method comprising: allocating an array of logical block arrays (LBAs) scatter lists. Additionally, the method may include: transmitting an initialization command to a target storage device, the initialization command providing the target storage device with an address pointing to the array of LBAs scatter lists. The method may also include: aggregating multiple read requests; and selecting an available LBA scatter list index. The method may further include: populating an LBA list indicated by the selected LBA scatter list index; and transmitting a subsequent command to the target storage device to cause the target storage device to calculate an address corresponding to a given entry in the selected LBA scatter list index.
[0013] In conjunction with any of the examples above, allocating the LBA scatter list array may include: setting the size of the LBA scatter list array, the maximum number of entries in the LBA scatter list array, and the number of lists in the LBA scatter list array.
[0014] In conjunction with any of the examples above, the method may also include: specifying the size of the offset.
[0015] Combining any of the examples above, populating the LBA list could include filling the LBA list with the multiple read requests.
[0016] Combining any of the examples above, the given entry in this LBA list could include START_OFFSET and END_OFFSET.
[0017] In conjunction with any of the examples above, the method may further include: receiving from the target storage device data corresponding to data requested in the given entry of the LBA scatter list index; and releasing the selected LBA scatter list index.
[0018] Combining any of the examples above, this subsequent command enables the target storage device to: calculate the LBA scatter list address based on the base host address and the LSLI index; and request a DMA transfer from host memory to the LBA scatter list shadow buffer.
[0019] Individually or in combination with any of the examples above, examples of this disclosure may include a method comprising: receiving an initialization command providing an address to an array of logical block arrays (LBAs). Additionally, the method may include: receiving a subsequent command providing an LBA scatter list index and a base host address. The method may also include: calculating an LBA scatter list address based on the base host address and the LBA scatter list index.
[0020] In conjunction with any of the examples above, the method may also include: saving the address of the LBA scatter list.
[0021] In conjunction with any of the examples above, the method may also include: allocating the total list size (TLS) of internal memory as a receive buffer for the LBA scatter list index.
[0022] In conjunction with any of the examples above, the method may also include: initiating a read command from a non-volatile medium for an entry in the LBA scatter list index.
[0023] In conjunction with any of the examples above, the method may also include: transmitting to the host computer data corresponding to the data requested in the entry of the LBA scatter list index.
[0024] In conjunction with any of the examples above, the method may also include: requesting a DMA transfer from host memory to the LBA scatter list shadow buffer. Attached Figure Description
[0025] The accompanying figures illustrate examples of the systems and methods.
[0026] Figure 1 An example system according to this disclosure is shown, the system including a host computer and a target storage device for calculating addresses corresponding to entries in the LBA scatter list using a logical block addressing (LBA) scatter list;
[0027] Figure 2 An example of a method for calculating an address corresponding to an entry in an LBA scatter list using a logical block addressing (LBA) scatter list, performed according to an example of this disclosure, is illustrated.
[0028] Figure 3 A more detailed method for calculating the address corresponding to an entry in the LBA scatter list using a logical block addressing (LBA) scatter list, as illustrated in the example of this disclosure, is given.
[0029] Figure 4 An example of a method for calculating the address corresponding to an entry in an LBA scatter list using a logical block addressing (LBA) scatter list, performed according to an example of this disclosure, is illustrated; and
[0030] Figure 5 A more detailed method for calculating the address corresponding to an entry in the LBA scatter list using an example of this disclosure is illustrated.
[0031] Reference numerals for any illustrated element appearing in multiple different figures have the same meaning in all figures, and any reference or discussion of any illustrated element in the context of any particular figure also applies to every other figure (if any) in which the same illustrated element is shown. Detailed Implementation
[0032] According to one aspect of the invention, a system and method are provided for calculating addresses corresponding to entries in an LBA scatter list using a Logical Block Addressing (LBA) scatter list. The disclosed system and method provide finer granularity in input / output (I / O) operations than 4K (the physical format of most non-volatile media, such as solid-state drives (SSDs)) and increase interconnect throughput (such as high-speed peripheral component interconnects (PCIe)). The granularity of data can differ from the sector or logical block size of the non-volatile media. One method for providing data granularity is to use a smaller block size. SSDs can maintain a logic-to-physical (L2P) table in non-volatile memory, where there are entries for a given LBA. Smaller block sizes result in an increased number of L2P entries, requiring more non-volatile memory and random access memory (RAM) for the L2P table and increasing design costs. Another method is to read from a byte offset of the LBA. The LBA offset is a multiple of a specified offset size starting from the beginning of the logical block. Software algorithms can be used to read multiple small data units from the LBA offset of the non-volatile storage device. The disclosed systems and methods allow for deeper queuing and parallel synchronization. These benefits are achieved by aggregating many small requests into a list and sending a single command to the storage device, resulting in reduced cost and increased throughput. Similarly, available interconnect bandwidth is saved by using a single response message for multiple media reads.
[0033] Figure 1 A system according to an example of this disclosure is illustrated, comprising a host computer and a target storage device for calculating addresses corresponding to entries in an LBA scatter list using a logical block addressing (LBA) scatter list. The host computer 102 typically includes: a central processing unit (CPU), general-purpose processor, special-purpose processor, microcontroller, programmable logic controller (PLC), digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic unit, discrete hardware component, other programmable device, or any combination thereof, designed to perform the functions disclosed herein; memory, such as random access memory; cache; storage device, such as a hard disk; and various input / outputs (not shown). The host computer 102 typically runs software including an operating system and one or more device drivers.
[0034] The target storage device 103 may include a storage controller 104 and non-volatile memory 106. The target storage device 103 may be any suitable storage medium using block storage. For example, the target storage device 103 may be a solid-state drive with non-volatile memory provided by a flash memory device.
[0035] The storage controller 104 may be a central processing unit (CPU), general-purpose processor, special-purpose processor, microcontroller, programmable logic controller (PLC), digital signal processor (DSP), application-specific integrated circuit (ASIC), graphics processing unit (GPU), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic unit, discrete hardware component, other programmable device, or any combination thereof, designed to perform the functions disclosed herein. The storage controller 104 may include memory, such as non-transitory memory, such as non-volatile random access memory (NVRAM), including machine-readable instructions that, when executed by the storage controller 104, cause the storage controller 104 to perform one or more actions as described herein. The storage controller 104 may also include a direct memory access (DMA) controller 105 for coordinating DMA transfer operations. DMA transfers can copy a range of memory from one device to another. The task of the storage controller 104 may generally be to control and access the non-volatile medium 106 based on read and write commands transmitted from the host computer 102.
[0036] In some examples, host computer 102 may be a graphics processing unit (GPU). GPU logic can tolerate deep queues and high latency. Therefore, packets of requests from GPU threads can arrive at target storage device 103 in batches. Additionally, the Non-Volatile Memory High Speed (NVMe) protocol specifies 64-byte requests and 16-byte responses. In the read direction, a 16-byte response may consume significant bandwidth. For example, if the request is 256 bytes, the overhead bandwidth of the response message would be 6%. By batching multiple read requests into a single command and response, interconnect usage is reduced, thereby increasing the available bandwidth for user data.
[0037] The SET_LBA_SCATTER command can be used to provide an address corresponding to an array of Logical Block Addressing (LBA) scatter lists in the memory space of host computer 102. For this example, the array may include, but is not limited to, the addressability of a list of up to 64K bytes (e.g., a list of 256 256-byte offsets). The SET_LBA_SCATTER command can specify the parameters OFFSET_SIZE, TOTAL_BYTES, MAX_SIZE_OF_LIST, and NUMBER_OF_ENTRIES. Target storage device 103 may accept the LBA scatter list retrieved from host computer 102 by the SET_LBA_SCATTER command and may create a matching array or a buffer managing target storage device 103 to delay the transfer of the LBA scatter list until target storage device 103 has slots available. Each LBA scatter list entry may have a size of 64 bits. The most significant byte (MSB) of the corresponding LBA scatter list entry may be START_OFFSET entering the logical block. START_OFFSET can be a multiple of OFFSET_SIZE specified in the SET_LBA_SCATTER command (maximum 256 entries). The next lower byte (8 bits) can be END_OFFSET. END_OFFSET can range from 0 to 255.
[0038] The READ_LBA_SCATTER command can be used to aggregate multiple read requests into a single command. The READ_LBA_SCATTER command can provide a 16-bit first parameter SCATTER_LBA_INDEX and a 16-bit second parameter NUMBER_OF_ENTRIES. Using the NVMe protocol, host computer 102 can provide a scatter-gather list (SGL) or physical region page (PRP) as the destination location of the scatter list of LBAs to be transferred in host memory. The target storage device 103 can transfer in a normal state when all data specified in the LBA scatter list has been copied to host memory, such as by using direct memory access (DMA). In the case of an unrecoverable read error (URE), the target storage device 103 can provide a state indicating the bitmask on the non-error segment, and host computer 102 can read the complete logical block.
[0039] In an example where the target storage device 103 is a Serial Attached Small Computer System Interface (SCSI) (SAS) or Serial Advanced Technology Attachment (SATA) drive, it will not have the ability to directly transfer data from host memory using DMA transfer. In this case, the host computer 102 may send a SET LBA-SCATTER LIST command to the target storage device 103 to set the LBA SCATTER_LIST_INDEX (LSLI), and in the DATA-OUT phase of the SCSI command (where the host computer 102 requests the target storage device 103 to transfer data to the host computer 102), the host computer 102 may send a list of LBA_SCATTER_LISTs. Upon completion of this command, the host computer 102 may use the LBA READ-SCATTER-LBA command, where the index number matches the LSLI set in the SET LBA-SCATTER LIST command. The target storage device 103 may provide static parameters to indicate the maximum number of supported LBA_SCATTER_LISTs and the OFFSET_SIZE.
[0040] Figure 2 An example of a method for calculating an address corresponding to an entry in an LBA scatter list using an example of this disclosure is illustrated. Method 200 can be implemented using a driver coupled to a processor or any other system capable of operating to implement method 200 on a host computer coupled to a block storage device, such as an SSD. Although examples have been described above, other variations and examples may be made according to this disclosure without departing from the spirit and scope of these disclosed examples.
[0041] At box 210, a storage controller driver running on host computer 102 may allocate an array of logical block arrays (LBAs). The LBA scatter list array may have a size of total list size (TLS) bytes. TLS may be equal to the maximum number of entries per list (MLS) multiplied by the size of a single entry, and then multiplied by the number of lists (NOL). For example, the size of a single entry may be eight bytes. After allocating the LBA scatter list array, the storage controller driver can identify a given LBA scatter list by the entries in the list index (LSLI). List index values may start from "0" and their maximum value is one less than the number of lists. In some examples, box 210 may occur during system or driver initialization.
[0042] At box 220, the storage controller driver can direct the input to the target storage controller (such as...). Figure 1The target storage controller 104 (shown) transmits an initialization command to provide the base address of the LBA scatter list array. The initialization command can be the SET_LBA_SCATTER command described above. The SET_LBA_SCATTER command can use the address of the LBA scatter list array as a data pointer and provides TLS, MLS, and NOL values. In response to receiving the SET_LBA_SCATTER command, the storage controller can store the address pointing to the LBA scatter list array in local memory and allocate TLS in internal memory as a receive buffer for the LBA scatter list. This receive buffer can be referred to as the LBA scatter list shadow buffer.
[0043] At box 230, the storage controller driver can aggregate multiple read requests. These read requests can be generated by an application executing on the host computer and can seek data stored in dispersed (e.g., random) locations on the target storage device. Box 230 can occur at runtime.
[0044] At box 240, the storage controller driver can select an available LBA scatter list index, which is referenced by a list index (LSLI). For example, four LBA scatter lists can be allocated in an array of LBA scatter lists. Two of the LBA scatter lists are in use, so at box 240, the storage controller driver can select one of the other two LBA scatter list indices.
[0045] At box 250, the storage controller driver can populate the LBA scatter list indicated by the selected LSLI (from box 240). The storage controller driver can populate the LBA scatter list based on the read requests aggregated at box 230. For a given read request in a pool of multiple read requests, the storage controller driver can populate a given entry in the LBA scatter list with the LBA, START_OFFSET, and END_OFFSET obtained from the address and range of the individual read request. A given entry in the LBA scatter list can correspond to a given read request. For example, an entry in the LBA scatter list can be a 64-bit value allocated as follows: 48 bits specify the LBA to be read, 8 bits specify the START_OFFSET (of OFS size), and 8 bits specify the END_OFFSET. END_OFFSET can be used to specify multiple adjacent offsets. For example, in a 4K block with a 256B offset, to read the entire block, START_OFFSET=0 and END_OFFSET=15.
[0046] Additionally, when the storage controller driver constructs the LBA scatter list, it can also construct a data pointer for the destination buffer in host memory. The data pointer can be contiguous memory or a scatter-cluster list (SGL).
[0047] At box 260, the storage controller driver may transmit a subsequent command to the target storage device to cause the target storage device to calculate the address corresponding to a given list in the selected LBA scatter list array. The subsequent command may be a READ_LBA_SCATTER command. The READ_LBA_SCATTER command may include the previously selected LBA scatter list index (LSLI), which may be the total data transfer bytes (TDTB) specified by the total number of bytes in all entries in the LBA scatter list, and the scatter list count (SLC) as the number of entries in the LBA scatter list.
[0048] The storage controller at the target storage device can receive the READ_LBA_SCATTER command. The storage controller can use the LBA Scatter List Index (LSLI) to calculate the address of the LBA scatter list based on the address stored by the storage controller during system initialization. The target storage device can then request that the LBA scatter list be transferred from the host memory DMA to the shadow LBA scatter list.
[0049] although Figure 2 A specific number of operations related to method 200 are disclosed, but method 200 is available in more... Figure 2 The number of operations described can be increased or decreased to perform the task. Furthermore, although... Figure 2 The specific order of operations to be performed in method 200 is disclosed, but the operations including method 200 can be performed in any suitable order.
[0050] Figure 3 A more detailed method for calculating the address corresponding to an entry in an LBA scatter list using an example performed according to this disclosure is illustrated. Method 300 can be implemented using a driver coupled to a processor or any other system capable of operating to implement method 300 on a host computer coupled to a block storage device such as an SSD. Although examples have been described above, other variations and examples may be made according to this disclosure without departing from the spirit and scope of these disclosed examples.
[0051] At box 310, the storage controller driver in the host computer can allocate an array of logical block array (LBA) scatter lists. After allocating the LBA scatter list array, the storage controller driver can identify a given LBA scatter list by entries in the list index (LSLI). List index values can start from "0" and their maximum value is one less than the number of lists. In some examples, box 310 may occur during system or driver initialization.
[0052] At box 312, the storage control circuitry can set the number of lists (NOL) in the LBA scatter list array.
[0053] At box 314, the storage controller driver can set the maximum number of entries per list. The maximum number of entries in the LBA scatter list array can be indicated by a value reflecting the maximum number of entries per list (MLS), i.e., the MLS value.
[0054] At box 316, the storage controller driver can set the size of the scatter list array. The array can have a size of the total list size (TLS) in bytes. The TLS can be equal to the MLS multiplied by the size of a single entry, and then multiplied by the NOL. For example, the size of a single entry can be eight bytes. Boxes 312, 314, and 316 can occur serially or concurrently and can be part of the allocation of the LBA scatter list array (box 310).
[0055] At box 318, the storage controller driver can specify the size of the offset (OFS) in bytes. The OFS can be a parameter included in the initialization command (e.g., the SET_LBA_SCATTER command) transmitted at box 320.
[0056] At box 320, the storage controller driver may transmit an initialization command to the target storage controller in the target storage device to provide the base address of the LBA scatter list array. The initialization command may be the SET_LBA_SCATTER command described above. The SET_LBA_SCATTER command may use the address of the LBA scatter list array as a data pointer and provide TLS, MLS, and NOL values. In response to receiving the SET_LBA_SCATTER command, the storage controller may store the address pointing to the LBA scatter list array in local memory at the target storage device and may allocate TLS in internal memory as a receive buffer for the LBA scatter list. This receive buffer may be referred to as the LBA scatter list shadow buffer.
[0057] At box 330, the storage controller driver can aggregate multiple read requests. These read requests can be generated by an application running on the host computer and can seek data stored in dispersed (e.g., random) locations on the target storage device. Box 330 can occur at runtime.
[0058] At box 340, the storage controller driver can select an available LBA scatter list index, which is referenced by an index (LSLI). For example, four LBA scatter lists can be allocated in an array of LBA scatter lists. Two of the LBA scatter lists are in use, so at box 340, the storage controller driver can select one of the other two LBA scatter lists.
[0059] At box 350, the storage controller driver can populate the LBA scatter list indicated by the selected LSLI (from box 340). The storage controller driver can populate the LBA scatter list based on the read requests aggregated at box 330. For a given read request in a pool of multiple read requests, the storage controller driver can populate a given entry in the LBA scatter list with LBA, START_OFFSET, and END_OFFSET. A given entry in the LBA scatter list can correspond to a given read request. For example, an entry in the LBA scatter list can be a 64-bit value allocated as follows: 48 bits specify the LBA to be read, 8 bits specify the START_OFFSET (of OFS size), and 8 bits specify the END_OFFSET.
[0060] Additionally, when the storage controller driver constructs the LBA scatter list, it can also construct a data pointer for the destination buffer in host memory. The data pointer can be contiguous memory or a scatter-cluster list (SGL).
[0061] At box 355, the storage controller driver can populate the LBA list with multiple read requests (aggregated at box 330), where a given entry in the LBA list includes START_OFFSET and END_OFFSET.
[0062] At box 360, the storage controller driver may transmit a subsequent command to the target storage device to cause the target storage device to calculate the address corresponding to a given list in the selected LBA scatter list array. The subsequent command may be a READ_LBA_SCATTER command. The READ_LBA_SCATTER command may include the previously selected LBA scatter list index (LSLI), which may be the total data transfer bytes (TDTB) specified by the total number of bytes in all entries in the LBA scatter list, and the scatter list count (SLC) as the number of entries in the LBA scatter list.
[0063] At box 370, the storage controller driver can receive data from the target storage device corresponding to data requested in a given entry of the LBA scatterlist index. The data requested in a given entry of the LBA scatterlist index can be from a given entry in the list populated at box 355. Specifically, the target storage device can use DMA transfer to transfer data from a given entry of the LBA scatterlist index (LSLI) to the host. The target storage device can transfer data to the host computer on an entry-by-entry basis, or the target storage device can temporarily store all data from the LBA scatterlist in internal memory and then initiate a single DMA transfer to host memory. The target storage device can send a single completion to the storage controller driver for the READ_LBA_SCATTER command.
[0064] At box 375, the storage controller driver can release the selected LBA scatter list index. By releasing the LBA scatter list index, it can be used with additional read requests (e.g., it can be used to select it at box 340). The storage controller driver can also notify the host computer that pending requests on the LBA scatter list have been completed and that the data is available in host memory.
[0065] although Figure 3 A specific number of operations related to method 300 are disclosed, but method 300 is available in more... Figure 3 The number of operations described can be increased or decreased to perform the task. Furthermore, although... Figure 3 The specific order of operations to be performed in method 300 is disclosed, but the operations including method 300 can be performed in any suitable order.
[0066] Figure 4 A method for calculating the address corresponding to an entry in an LBA scatter list using a logical block addressing (LBA) scatter list, as illustrated in the examples of this disclosure, is described. Method 400 can be implemented using a storage controller on a block storage device (such as an SSD) in conjunction with a processor or any other system capable of operating to implement method 400. Although examples have been described above, other variations and examples may be made according to this disclosure without departing from the spirit and scope of these disclosed examples.
[0067] At box 410, the storage controller may receive an initialization command from the storage controller driver on the host computer, thereby providing the base address of the LBA scatter list array. The initialization command may be the SET_LBA_SCATTER command described above. The SET_LBA_SCATTER command may use the address of the LBA scatter list array as a data pointer and provide TLS, MLS, and NOL values.
[0068] At box 420, the storage controller may receive a subsequent command from the storage controller driver to cause the storage controller to calculate the address corresponding to a given list in the selected LBA scatter list array. The subsequent command may be a READ_LBA_SCATTER command. The READ_LBA_SCATTER command may include previously (e.g., in...) Figure 2 In box 240 or at Figure 3 The LBA Distributed List Index (LSLI) selected in box 340 can be the total data transfer bytes (TDTB) which is the total number of bytes specified by all entries in the LBA Distributed List, and the Distributed List Count (SLC) which is the number of entries in the LBA Distributed List.
[0069] At box 430, the storage controller can calculate the LBA scatterlist address based on the base host address and the LSLI index. The LBA scatterlist address is the base address plus the LSLI multiplied by the MLS and 8 (the number of bytes in the entry). The storage controller can use the LBA scatterlist index (LSLI) to calculate the host address based on the address stored by the target storage device during system initialization.
[0070] although Figure 4 A specific number of operations related to method 400 are disclosed, but method 400 is available in more... Figure 4 The number of operations described can be increased or decreased to perform the task. Furthermore, although... Figure 4 The specific order of operations to be performed in method 400 is disclosed, but the operations including method 400 can be performed in any suitable order.
[0071] Figure 5 A more detailed method for calculating the address corresponding to an entry in an LBA scatter list using an example performed according to this disclosure is illustrated. Method 500 can be implemented using a storage controller on a block storage device (such as an SSD) in conjunction with a processor or any other system capable of operating to implement method 500. Although examples have been described above, other variations and examples may be made according to this disclosure without departing from the spirit and scope of these disclosed examples.
[0072] At box 510, the storage controller may receive an initialization command from the storage controller driver on the host computer, thereby providing the base address of the LBA scatter list array. The initialization command may be the SET_LBA_SCATTER command described above. The SET_LBA_SCATTER command may use the address of the LBA scatter list array as a data pointer and provide TLS, MLS, and NOL values.
[0073] At box 512, in response to receiving the SET_LBA_SCATTER command, the storage controller may save the address pointing to the LBA scatter list array in the local memory at the target storage device.
[0074] At box 514, the storage controller can allocate TLS in internal memory as a receive buffer for the LBA scatter list based on the SET_LBA_SCATTER command. The receive buffer may be referred to as a shadow LBA scatter list.
[0075] At box 520, the storage controller may receive a subsequent command from the storage controller driver to cause the storage controller to calculate the address corresponding to a given list in the selected LBA scatter list array. The subsequent command may be a READ_LBA_SCATTER command. The READ_LBA_SCATTER command may include previously (e.g., in...) Figure 2 In box 240 or at Figure 3 The LBA Distributed List Index (LSLI) selected in box 340 can be the total data transfer bytes (TDTB) which is the total number of bytes specified by all entries in the LBA Distributed List, and the Distributed List Count (SLC) which is the number of entries in the LBA Distributed List.
[0076] At box 530, the storage controller can calculate the LBA scatterlist address based on the base host address and the LSLI index. The LBA scatterlist address is the base address plus the LSLI multiplied by the MLS and 8 (the number of bytes in the entry). The storage controller can use the LBA scatterlist index (LSLI) to calculate the host address based on the address stored by the target storage device during system initialization.
[0077] At box 540, the storage controller may request the transfer of the LBA scatter list from the host memory DMA to the shadow LBA scatter list.
[0078] At box 550, after the DMA transfer copies the LBA scatter list to the shadow LBA scatter list, the storage controller can initiate a read command from the non-volatile medium for entries in the LBA scatter list.
[0079] At box 560, the storage controller can transfer data to the host computer corresponding to data requested in a given entry of the LBA scatterlist index. Specifically, the storage controller can use DMA transfer to transfer data from a given entry of the LBA scatterlist index (LSLI) to the host. The storage controller can transfer data to the host computer on an entry-by-entry basis, or the storage controller can temporarily store all data from the LBA scatterlist in internal memory and then initiate a single DMA transfer to host memory. The storage controller can also send a single completion to the storage controller driver for the READ_LBA_SCATTER command.
[0080] although Figure 5 A specific number of operations related to method 500 are disclosed, but method 500 is available in more... Figure 5 The number of operations described can be increased or decreased to perform the task. Furthermore, although... Figure 5 The specific order of operations to be performed in method 500 is disclosed, but the operations including method 500 can be performed in any suitable order.
[0081] Although examples have been described above, other variations and examples may be made in accordance with this disclosure without departing from the substance and scope of these disclosed examples.
Claims
1. An article comprising: Non-transitory memory, the non-transitory memory including machine-readable instructions, which, when executed by a processor, cause the processor to: Allocation Logical Block Array (LBA) scatter list array; An initialization command is sent to the target storage device, the initialization command being used to provide the target storage device with an address pointing to the LBA scatter list array; Aggregate multiple read requests; Select the available LBA scatter list index; Populate the LBA list indicated by the selected LBA scatter list index; as well as Subsequent commands are transmitted to the target storage device to cause the target storage device to calculate the address corresponding to a given entry in the selected LBA scatter list index.
2. The article of claim 1, wherein allocating the LBA scatter list array includes setting the size of the LBA scatter list array, the maximum number of entries in the LBA scatter list array, and the number of lists in the LBA scatter list array.
3. The article of manufacture according to any one of claims 1 to 2, wherein the processor is used to specify the size of the offset.
4. The article of manufacture according to any one of claims 1 to 3, wherein filling the LBA list comprises filling the LBA list with the plurality of read requests.
5. The article of manufacture according to any one of claims 1 to 4, wherein the given entry in the LBA list includes START_OFFSET and END_OFFSET.
6. The article of manufacture according to any one of claims 1 to 5, wherein the processor is used for: Receive data from the target storage device corresponding to the data requested in the given entry of the LBA scatter list index; and Release the selected LBA scatter list index.
7. The article of manufacture according to any one of claims 1 to 6, wherein the subsequent command causes the target storage device to: Calculate the LBA scatter list address based on the base host address and the LSLI index; and Request a DMA transfer from host memory to the LBA scatter list shadow buffer.
8. A method, the method comprising: Allocation Logical Block Array (LBA) scatter list array; An initialization command is sent to the target storage device, the initialization command being used to provide the target storage device with an address pointing to the LBA scatter list array; Aggregate multiple read requests; Select the available LBA scatter list index; Populate the LBA list indicated by the selected LBA scatter list index; as well as Subsequent commands are transmitted to the target storage device to cause the target storage device to calculate the address corresponding to a given entry in the selected LBA scatter list index.
9. The method of claim 8, wherein allocating the LBA scatter list array includes setting the size of the LBA scatter list array, the maximum number of entries in the LBA scatter list array, and the number of lists in the LBA scatter list array.
10. The method according to any one of claims 8 to 9, the method comprising: Specify the size of the offset.
11. The method according to any one of claims 8 to 10, wherein populating the LBA list comprises populating the LBA list with the plurality of read requests.
12. The method according to any one of claims 8 to 11, wherein the given entry in the LBA list includes START_OFFSET and END_OFFSET.
13. The method according to any one of claims 8 to 12, the method comprising: Receive data from the target storage device corresponding to the data requested in the given entry of the LBA scatter list index; as well as Release the selected LBA scatter list index.
14. The method according to any one of claims 8 to 13, wherein the subsequent command causes the target storage device to: Calculate the LBA scatter list address based on the base host address and the LSLI index; and Request a DMA transfer from host memory to the LBA scatter list shadow buffer.
15. A method comprising: Receives an initialization command that provides an address to an array of logical block arrays (LBAs) scatter list arrays; Receive subsequent commands providing the LBA scatter list index and base host address; as well as The LBA scatter list address is calculated based on the base host address and the LBA scatter list index.
16. The method of claim 15, wherein the method comprises: Save the address of the LBA scatter list.
17. The method according to any one of claims 15 to 16, the method comprising: Allocate the total list size (TLS) of internal memory as a receive buffer for the LBA scatter list index.
18. The method according to any one of claims 15 to 17, the method comprising: A read command is initiated from the non-volatile medium for each entry in the LBA scatter list index.
19. The method according to any one of claims 15 to 18, the method comprising: Transmit data to the host computer that corresponds to the data requested in the entry of the LBA Distributed List Index.
20. The method according to any one of claims 15 to 19, the method comprising: Request a DMA transfer from host memory to the LBA scatter list shadow buffer.
