DE102020206596A1 - Method for operating a storage medium and computing unit with storage medium - Google Patents

Abstract

The invention relates to a method for operating a storage medium (120) which provides a non-volatile memory that can be written to without prior erasure and on which data (115) can be stored in accordance with a physical address space (121), with data (115) in the Memory is written and / or read from the memory by addressing a virtual address space (141) which is mapped into the physical address space (121) according to a changeable assignment (130), with physical sub-areas (122 ) are provided, to which virtual sub-areas (142), which are addressed via the virtual address space (141), are assigned according to the changeable assignment (130), and the assignment (130) is repeatedly changed and the data is retained while maintaining it Position in the virtual sub-areas (142), are transferred to other physical sub-areas (122), as well as a computing unit (100) with a such storage medium (120).

Description

The present invention relates to a method for operating a storage medium which provides a non-volatile memory, as well as a computing unit with such a storage medium and a computer program for its implementation.

State of the art

Computing units such as control units in vehicles (i.e. in the automotive sector) generally not only have to store a program code, but also - during the runtime of the software running on them - data in a non-volatile manner. EEPROMs or flash memories are traditionally used as hardware for this purpose. The functionality of the associated software is standardized by the so-called AUTOSAR consortium. However, these standards only define the functionality of the software, but not the way in which the data is organized on the storage medium.

Disclosure of the invention

According to the invention, a method for operating a storage medium as well as a computing unit and a computer program for its implementation are proposed with the features of the independent patent claims. Advantageous refinements are the subject matter of the subclaims and the description below.

The invention is concerned with a method for operating a storage medium which provides a non-volatile memory on which data can be stored in accordance with a physical address space. Such a method can in particular be carried out on a computing unit such as a control unit in a vehicle - then with the appropriate software. In this respect, the invention also relates to a corresponding computing unit with such a storage medium.

When writing to such a storage medium, several points must usually be taken into account. Every erasure or programming process places a physical load on the memory cell (s). The number of permitted erasing or programming processes per memory cell of the memory technologies used is therefore limited. Beyond the load limit, robust operation can no longer be guaranteed due to wear and tear.

The non-volatile data differ not only in terms of their content, but also in particular in terms of the frequency with which they are stored. The storage frequency of individual data often exceeds the number of permitted deletion or programming processes. In order to reduce the resulting load on the storage medium, the deletion and programming processes must be distributed over the medium (so-called "wear leveling").

Individual memory cells or larger, contiguous memory areas can fail due to hardware defects. No data may be stored in such memory cells or memory areas. Flash memory (and also other storage media with non-volatile memory) cannot be overwritten at will and must be deleted in larger areas (in so-called flash banks) before programming.

Previous options for dealing with such storage media, especially flash memories, are sometimes associated with disadvantages, as will be briefly explained below.

Associated data are summarized and managed in the form of blocks. If a data block has changed and there is a write job for it, the block can be appended to the end of the occupied area, i.e. to the next free physical address. The storage medium will fill up over time due to a large number of write jobs. This increases the times it takes to find blocks. Before this state is reached, the flash bank with the oldest data should be deleted to make space for new write jobs. It may be necessary to move data from this bank to the end of the occupied area in order not to get lost. Both the deletion of a bank and the search for and relocation of data is time-consuming and slows down or blocks the processing of write or read jobs.

If a memory cell or a memory area can no longer be erased and / or programmed due to a hardware defect, the data can be placed behind the defective area, i.e. at the next error-free physical address. The unsuccessful programming attempts will take a long time, especially in the case of larger defective areas (e.g. a defective wordline).

Against this background, it is now proposed that data be written to the memory and / or read from the memory by addressing a virtual address space that is mapped into the physical address space according to a changeable assignment (a “mapping”).

The storage medium in a computing unit such as a control unit is normally used by the microcontroller or processor (as part of the computing unit) via the physical address space addressed. In the context of the present invention, this physical memory or address space is abstracted into a virtual (or logical) memory or address space. If data is written to the physical memory (or its address space) or read from it again, this is now done with the interposition of a virtual or logical address space. Between the physical and logical address space there is a dynamic mapping (assignment) that can be changed during the runtime of the software, whereby physical sub-areas of the storage medium are assigned virtual sub-areas that are addressed via the virtual address space. This assignment is stored in particular, for example also redundantly, in the physical subareas, so the virtual address or a characteristic value for it, such as an ID, is also stored at the assigned physical address. In the course of the invention, while maintaining their position in the virtual sub-areas, the data are then repeatedly transferred to other (free) physical sub-areas and the mapping is adapted accordingly. In other words, the data is copied over while the virtual address to be addressed is retained.

This has various advantages. In this way, defective areas in the physical memory can be mapped very easily to another area or assigned to another area, i.e. the actual defective area is no longer assigned to a virtual area; Rather, such a virtual area is then assigned to a physical area (previously in particular free or kept as a reserve). A time-consuming search for the end of the defective area is no longer necessary.

Since the data remain permanently at the same virtual addresses and the mapping to the physical addresses is known, there is no need for a time-consuming search in the storage medium. There is also no need to always append data (blocks) to the end of the occupied area or to move to another bank.

The repeated, especially regular, copying of the data (with a change in the assignment) also prevents rarely programmed cells (physical subareas) from drifting away and thus prevents data loss. All data are regularly reprogrammed physically using the copying or transferring mechanism. This counteracts charge losses that would naturally lead to loss of data after years.

Even if the invention can preferably be used with EEPROM, PCM (Phase-Change Random Access Memory) or MRAM (Magnetoresistive Random Access Memory), it basically also works with other non-volatile memories. The invention has particular advantages in memories that can be written to without prior erasure, since the physical sub-areas can then be transferred in a targeted manner. When saving with areas that can only be overwritten after previous deletion, such as conventional flash memories, the entire area (e.g. a so-called flash bank or sector) must first be deleted before a single cell can be programmed. This generally erases a much larger area than is then programmed, which means too much wear and tear for the essentially relevant applications here, such as vehicle control units.

The advantages mentioned can be achieved in particular if the mapping or the assignment is always changed when at least one criterion is met, for example each time the computing unit is started and / or after a predetermined amount of data into the memory and / or the sub-area written data and / or a predetermined number of write operations of the memory and / or the sub-area.

In addition to the mapping, a sequence counter (as administrative information) is preferably also stored in the physical sub-areas. This can be used to identify the age of a sub-area. Each time data is written to a previously unused sub-area, it is newly created and assigned (mapped) accordingly. Each new sub-area can, for example, receive an incremented sequence counter as a further part of the management information, i.e. the larger the value, the newer the sub-area.

Preferably, however, a physical sub-area is or is only assigned to a virtual sub-area if data is or will also be stored in the virtual sub-area. Virtual sub-areas that have not yet been used are then not assigned any physical sub-areas. As a result, a reserve of physical sub-areas can be made available, on the one hand to be able to rotate the oldest physical sub-area to a different location and on the other hand to be able to remap (ie reassign) defective physical sub-areas. The greater the reserve, the greater the scope for action in the event of defects. In particular, the physical subareas are stored in a contiguous manner (generally sorted according to the sequence counter), as this significantly simplifies and accelerates the search for the oldest and newest physical subarea. This also ensures that read and write access is not arbitrary via the memory must be distributed, but relate to contiguous memory areas.

A computing unit according to the invention, e.g. a control unit of a vehicle, is set up, in particular in terms of programming, to carry out a method according to the invention.

The implementation of a method according to the invention in the form of a computer program or computer program product with program code for performing all method steps is advantageous, since this causes particularly low costs, especially if an executing control device is used for other tasks and is therefore available anyway. Suitable data carriers for providing the computer program are in particular magnetic, optical and electrical memories, such as in particular PCM or MRAM, but also, for example, hard drives, flash memories, EEPROMs, DVDs, etc. A program can also be downloaded via computer networks (Internet, intranet, etc.) .

Further advantages and embodiments of the invention emerge from the description and the accompanying drawing.

The invention is shown schematically in the drawing using an exemplary embodiment and is described below with reference to the drawing.

Figure list

• 1 shows schematically a computing unit according to the invention in a preferred embodiment.
• 2 and 3 show various partial sequences of a method according to the invention in a preferred embodiment.

Embodiment (s) of the invention

In 1 is a schematic of a computing unit according to the invention 100 in a preferred embodiment, for example a control unit of a vehicle, with a storage medium 120 with non-volatile memory, e.g. PCM memory, and a microcontroller 110 , shown. The following is the mode of operation of this arithmetic unit and thus also of the method according to the invention for the data 115 read from or written to memory.

The physical address space 121 of the storage medium 120 becomes, for example, evenly in physical sub-areas 122 which are also referred to as physical rows or rows in the following. The data 115 (or data blocks) that have to be managed (ie read and written) by the control unit are stored in a virtual memory 140 with a virtual address space 141 placed, which is also evenly in virtual or logical rows 142 is divided. The assignment 130 (also known as "mapping") between virtual rows 142 and physical series 122 is used as management information in the physical ranks 122 , and there especially in the head areas 123 ("Header") and / or footer areas 124 ("Footer") or here redundantly in both, stored. In addition, this assignment can also be present in the executing software. The first bytes of a row can be designated as the header area and the last bytes as the footer area. A row then consists of a user data area and a header area and / or a footer area. As mentioned, the data in the header area and footer area can be identical.

A physical series 122 cells that can no longer be written to due to a hardware defect are marked as defective in the management information and are therefore excluded from mapping, as in the example shown, the last row (seen from above, i.e. the bottom) 122 .

In general, the storage medium is 120 not completely filled. Free sub-areas, ie unused physical rows 122 , are marked as reserve in the management information and can thus be released for changes (e.g. rotation of the rows) or extensions in the mapping (saving new data), as in the example shown, the fifth and sixth rows (seen from above) 122 (these are not assigned to any virtual series).

• Physikalischer Adressraum: 0x1C80000 - 0x1C87FFFF Größe des physikalischen Speichers: 512kB
• Gewählte Größe einer physikalischen Reihe: 1024B (Vielfaches der Bits auf einer Wordline)
• Größe der Verwaltungsinformation (Header, Footer) je Reihe: 2 * 16B = 32B

An example of such an assignment or such a mapping is explained below, with the following assumptions:

• Physical address space: 0x1C80000 - 0x1C87FFFF Size of the physical memory: 512kB
• Selected size of a physical row: 1024B (multiple of the bits on a wordline)
• Size of the management information (header, footer) per row: 2 * 16B = 32B

It should be noted here that, given the size of the physical row, the multiple of the bits on a wordline is important insofar as it allows errors on individual wordlines to be adequately dealt with.

• Anzahl an Reihen: 512kB / 1kB = 512
• Größe (Nutzdaten) einer logischen (virtuellen) Reihe: 1024B - 32B = 992B Verwaltungsinformationen insgesamt: 512 * 32B = 16kB
• Größe des virtuellen Speichers insgesamt: 512kB - 16kB = 496kB = 992B * 512 Adressraum des virtuellen Speichers: 0x00000 - 0x7BFFF

This results in:

• Number of rows: 512kB / 1kB = 512
• Size (user data) of a logical (virtual) row: 1024B - 32B = 992B Total administrative information: 512 * 32B = 16kB
• Total size of the virtual memory: 512kB - 16kB = 496kB = 992B * 512 Address space of the virtual memory: 0x00000 - 0x7BFFF

1. 1. Virtuelle Adresse 0x00000: 0x00000 / 992 = 0, Rest 0 Daten 210 werden in die logische Reihe ID 000 mit Offset 0 gelegt.
2. 2. Virtuelle Adresse 0x3DF80: 0x3DF80 / 992 = 255, Rest 864 Daten 210 werden in die logische Reihe ID 255 mit Offset 864 gelegt.
3. 3. Virtuelle Adresse 0x3E000: 0x3E000 / 992 = 256, Rest 0 Daten 210 werden in die logische Reihe ID 256 mit Offset 0 gelegt.
4. 4. Virtuelle Adresse 0x7BF80: 0x7BF80 / 992 = 511, Rest 864 Daten 210 werden in die logische Reihe ID 511 mit Offset 0 gelegt.

In the empty storage medium, as in 2 illustrated, 128B data is now written to the following addresses in the order given:

1. 1. Virtual address 0x00000: 0x00000 / 992 = 0, remainder 0 data 210 are placed in the logical row ID 000 with offset 0.
2. 2. Virtual address 0x3DF80: 0x3DF80 / 992 = 255, remainder 864 data 210 are in the logical series ID 255 with offset 864 placed.
3. 3. Virtual address 0x3E000: 0x3E000 / 992 = 256, remainder 0 data 210 are in the logical series ID 256 placed with offset 0.
4. 4. Virtual address 0x7BF80: 0x7BF80 / 992 = 511, remainder 864 data 210 are in the logical series ID 511 placed with offset 0.

Since no rows were stored in the storage medium at the beginning, the new rows are physically placed at the next free address. Accordingly, there is no fixed assignment between physical and logical address, but this assignment results from 10. The ID here represents part of the administrative information.

This results in the in 2 memory image shown 200 . In the example there are still 508 unused rows (labeled #FREE) available as a reserve. Conversely, the ID of the logical row and the offset of the data within the row can always be used to clearly identify the virtual address to which the data belongs.

In addition to the mapping, according to one embodiment, a sequence counter (referred to here as cnt) is also stored as administrative information. This can be used to identify the age of a series. Each time data is written to a previously unused row, it is newly created and assigned (mapped) accordingly, i.e. provided with the unique ID that matches the virtual address as part of the administrative information. Each new row receives an incremented sequence counter as a further part of the management information, i.e. the larger the value, the newer the row.

Each time data is written to a row that is already in use, the corresponding area is overwritten, the ID and sequence counter are retained. Frequent writing in the same (physical) rows would put a lot of strain on them. To avoid this, too, the rows are rotated repeatedly on the storage medium, in particular whenever one or more criteria are met. Such criteria are, for example: The amount of programmed data (volatile in RAM) is counted for each programming job. If a configured limit value is exceeded, a rotation is triggered and the counter is reset to zero. The limit value can be, for example, 30 times a physical row, i.e. 30 kB for a physical row length of 1 kB. In systems in which only very little data is programmed, a rotation can alternatively or additionally be forced with every first programming job and / or after every first start-up of the computing unit - regardless of the counter. Forcing the rotation when writing for the first time ensures that it is rotated even when the write load is low. If only a minimum number of written bytes were to be rotated, it could happen that the minimum number is not reached in each individual power-on cycle and is therefore never rotated. By forcing a rotation when writing for the first time, it is rotated safely and thus an even load is enforced.

In the in 3 The two examples shown (above and below), a rotation is carried out, for example, as follows: Five of the eight existing logical rows are occupied, with sequence counters cnt = 0 (oldest) to cnt = 4 (newest). Every two rows are rotated. The initial state before the rotation is shown on the left, while the state after the rotation is shown on the right. The oldest and newest rows are relevant for a rotation and are marked with arrows. The oldest row is moved behind the newest. The dashed arrows stand for the oldest (top) or newest (bottom) row after the first rotation; they thus represent an intermediate state between the left (initial state) and right image (state after two rotations). In both cases, the two oldest rows, ie those with the smallest counter value, are rotated. Since the next free row is always used when rows are rewritten, a continuous free area generally remains (as long as no defective row has been sorted out in between).

The "wear leveling" can be improved even further, e.g. by not only rotating the logical rows physically, but also the data within a row. For example, a logical row can be viewed as a ring buffer and the data it contains can be shifted by a fixed offset at certain (time) intervals or regularly (a criterion here would be, for example, the amount of data programmed into a row), which is then stored as part of the management information. The offset then designates the physical address 0x0 in this row.

Claims (12)

A method for operating a storage medium (120) which provides a non-volatile memory on which data (115) can be stored in accordance with a physical address space (121), wherein data (115) are written into the memory and / or read from the memory by addressing a virtual address space (141) which is mapped into the physical address space (121) according to a changeable assignment (130), physical partial areas (122) being provided on the storage medium (120) to which virtual partial areas (142), which are addressed via the virtual address space (141), are assigned according to the changeable assignment (130), and the assignment (130) being repeatedly changed and the data being transferred to other physical subareas (122) while maintaining their position in the virtual subareas (142). Procedure according to Claim 1 , the assignment (130) being stored in the physical sub-areas (122). Procedure according to Claim 1 or 2 wherein, if there is a defect in a physical sub-area, the assignment is changed and the physical sub-area (122) with the defect according to the assignment (130) is not assigned to a virtual sub-area. Method according to one of the preceding claims, wherein a physical sub-area (122) is only assigned to a virtual sub-area (142) if data (115) are stored in the virtual sub-area (142). Method according to one of the preceding claims, wherein one or more physical sub-areas (122) according to the assignment (130), at least initially, are not assigned to a respective virtual sub-area and are kept free as a reserve. Method according to one of the preceding claims, wherein the assignment (130) is changed if at least one criterion is met. Procedure according to Claim 6 , wherein the at least one criterion is starting a computing unit (100) comprising the storage medium (120) and / or a predetermined amount of data written into the memory and / or the sub-area and / or a predetermined number of write operations to the memory and / or of the sub-area. Method according to one of the preceding claims, wherein an EEPROM, a phase change memory or a magnetoresistive random access memory is used as the storage medium (120). Computing unit (100) with a storage medium (120), which is set up to carry out all method steps of a method according to one of the preceding claims. Arithmetic unit (100) Claim 9 which is designed as a control device, in particular as a control device of a vehicle. Computer program that causes a processing unit (100) to perform all method steps of a method according to one of the Claims 1 until 8th to be carried out when it is executed on the computing unit (100). Machine-readable storage medium with a computer program stored thereon Claim 11 .

Images