CA1176377A - Guest architectural support in a computer system - Google Patents

Guest architectural support in a computer system

Info

Publication number
CA1176377A
CA1176377A CA000400593A CA400593A CA1176377A CA 1176377 A CA1176377 A CA 1176377A CA 000400593 A CA000400593 A CA 000400593A CA 400593 A CA400593 A CA 400593A CA 1176377 A CA1176377 A CA 1176377A
Authority
CA
Canada
Prior art keywords
guest
tlb
host
address
translation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired
Application number
CA000400593A
Other languages
French (fr)
Inventor
Robert J. Bullions, Iii
Thomas O. Curlee, Iii
Peter H. Gum
Bruce L. Mcgilvray
Ethel L. Richardson
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
International Business Machines Corp
Original Assignee
International Business Machines Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by International Business Machines Corp filed Critical International Business Machines Corp
Application granted granted Critical
Publication of CA1176377A publication Critical patent/CA1176377A/en
Expired legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F12/00Accessing, addressing or allocating within memory systems or architectures
    • G06F12/02Addressing or allocation; Relocation
    • G06F12/08Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
    • G06F12/10Address translation
    • G06F12/1027Address translation using associative or pseudo-associative address translation means, e.g. translation look-aside buffer [TLB]
    • G06F12/1045Address translation using associative or pseudo-associative address translation means, e.g. translation look-aside buffer [TLB] associated with a data cache
    • G06F12/1063Address translation using associative or pseudo-associative address translation means, e.g. translation look-aside buffer [TLB] associated with a data cache the data cache being concurrently virtually addressed
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/45566Nested virtual machines
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44Arrangements for executing specific programs
    • G06F9/455Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533Hypervisors; Virtual machine monitors
    • G06F9/45558Hypervisor-specific management and integration aspects
    • G06F2009/45583Memory management, e.g. access or allocation

Abstract

ABSTRACT OF THE DISCLOSURE

The described embodiment provides translation look-aside buffer (TLB) hardware in a CP that receives the results of double-level address translations to eliminate the need for having shadow tables for the second-level in a virtual machine (VM) environment.
Each TLB entry contains hardware which indicates whether the address sent by the CP Instruction Execution (IE) unit for translation is a guest or host/native request, and for a guest request if it is a real or virtual address. Intermediate translations for a double-level translation are inhibited from being loaded into the TLB. Guest entries are purged from the TLB without disturbing any host entries. An accelerated preferred guest mode in the CP forces its hardware adder translation hard ware to translate each accelerated preferred guest request, since it requires only a single level translation. A nonaccelerated guest request is instead translated by microcode. A
limit check register is provided to check preferred guest addresses without causing performance degradation.

Description

~7~3~7/7 GUEST ARCHITECTURAL SUPPORT IN A COMPUTER SYSTEM

A. _Field Of The Invention This specification generally relates to efficient emulation of one or more guest architectures in a multiprogrammable computer system to improve the per formance of the system when executing plural levels of operating systems using plural levels of architecture.
This invention obtains high performance operation of programs using guest computer architectures which may be the same as, or different from, the native architec ture for which the data processing system is designed to obtain its most efficient execution of programs (i.e.
best performance) without using emulation support.

This invention is particularly oriented to improving the performance of a virtual machine (VM) system using plural levels of architectures in which each architecture level is capable of using virtual addressing with dynamic address translation ~DAT) with the resultant capability of having a user storage request require plural levels of DAT.

B. ~escri~tion Of The Prior Art Many different computer architectures (e.g. IBM*
System/370, System/360~ 7090, 1401, etc.) exist, each having substantial libraries of programs involving large monetary investments. On the other hand, any parti cular computer system has its optimum performance tailored to one hardware architecture, which is generally called its native architecture. Conse~uently, conventional com puter systems which include modi~ications for emulating nonnative architectures run the programs written for the nonnative architectures less efficiently than *Trade Mark a ~

3'7~
2--programs running on the native architecture. ~ever-theless, emulation means has been provided in the prior art on many prior commercial computer systems to enable a particular system to execute programs re~uiring a given foreign hardware architecture, even though such non-native architecture may ru~ less efficiently ~e.g. 1401 on a System/360 model), in order to avoid the high cost of reprogramming such existing programs to the native architecture of a particular computer system.

The situation is further complicated by having different operating systems designed to run on the same hardware architecture (e.g. VM/370, MVS/370, DOS/
370, etc.), which cause operational incompatibilities among programs written to operate under the different operating systems, even though they operate under the same hardware architecture (e.g. use of the same set of machine instructions and machine facilities). For example, a program written to operate under DOS/370 may not ne~essarily be operational under MVS/370, even though both programs use the machine instructions of the S/370 architecture. One reason is that application programs often rely on special control programming facilities, e.g. program inclusion of special macro-instructions, job control language statements, etc.,provided by a particular operating system to ease the program design burden, but the special facilities may make the program dependent upon the particular operating system.

The situation is further complicated by the economic need to share an expensive computer system among a large number of users in a manner which makes the computer system appear to each user as if that user has all (or a substantial part) of the facilities , .
3~ 7 of the system. This problem has foun~ a solution in current commercial interactive multiprogramming systems such as the IsM V~1/370 and MVS/370 (using TSO) operating systems. The existin~ ~/370 system has the advantage of being able to simulate (by its software) multilevels of architecture, so that its users can concurrently and relatively efficiently execute programs written for any of the different operating systems (e.g. MVS, DOS, VM, etc.) which use the System/370 or Sys-tem/360 hardware architecture.

A further complicating factor is the virtual address translation found in a hardware architecture, such as IBM System/370 which provides a virtual addressing architecture to simplify user pro~ram wr~ting by eliminating the need for the user to control the use of real storage space, as is essential in non-virtual macnines, i.e. System/360 or System/370 with DAT off.

Further difficulty is caused when two levels o~ DAT
architecture are used. This occurs, for example, when a V~I/370 system permits a user program to execute on one operating system (e.g. MVS/370) running on a second level System/370 architecture which i5 being emulated under a first level System/370 architecture running another operating system (e.g. VM/CP), in which both operating systems have DAT on to require two levels of address translation to translate each address requested by the user program in order to access the requested data in main storage. Such double-level translation increases the translation overheadin comparison to running the same user program under only a single level of operation system. For example, the double-level translation of a single request Po9-80-005 ~'7~

involves eight accesses of segment and page tables in main storage, in comparison to a single-le~el trans-lation ~hich involves only two main storage accesses (i.e. one segment table entry and one page table entry) Since the System/370 hardware on which the VM/370 System was run did not provide for double~level address translation, software constructed shadow translation tables when a second level operating system was using DAT. Shadow translation tables provided a me~hod to allow the System/370 hardware to translate second level virtual addresses by using a standard two-reference translation. If a page fault, or segment fault occurred in the shadow translation tables, software would be notified, by a standard DAT fault, of the condition, and software would fill in the correct shadow table entry by performing a double-level address translation. There was one shadow translation table constructed for each user which was using DAT.

The use of System/370 dynamic address translation architecture by the VM/370 control program to represent two levels of translation is described in "~/370 Data Areas and Control Block Logic" Form. N. SY20-0884, and in "VM/370: System Logic and Problem Determination Guide, Volume 1 (CP Component)" Form ~o. SY20-0886.

Each shadow table uses the standard format of segment and page tables. The shadow table for each user receives the results of each double-level translation by the guest (i.e. the corresponding page frame absolute address for the requested virtual address), and there-after allows the next request by the same user to the same page frame in main storage to be obtained with only a single-level translation, as long as that entry is not 7~

purged from the shadow table. The shadow tables for a user include a shadow segment table and one or more shadow page tables in main storage. The shadow tables are used like a translation look-aside buffer (TL~).
Each user request is first made to the user's shadow table, because if the request is found in the shadow table only two main storage accesses were required to obtain the resultant main storage address for the user requested data. Only if the user requested address was not found in his shadow tables did the software need to perform double-level translation, then put the resultant translation into the required shadow table entry for subsequent use.

Therefore, the shadow tables reduce the number of lS main storage accesses from eight to two for the second and later requests for the address of the same page frame, as long as the entry is not purged from the user's shadow table. The shadow table TLB is effectively an extension of the hardware TLB found in the commercial large IB~ CPU's. The hardware TLB (also called DI~T) does not require any main storage access when obtaining the address for additional requests to the same page frame. Unfortunately, the prior DLATs could only contain single-level address translations communicated by the single-level hardware DAT in the CPU. Thus, a request found in a shadow table by means of a single-level DAT translation was then put into the DLAT, so that the third request to a page frame would obtain its translated address from the DLAT without further main storage accesses. However, some of the pre-ceeding main storage accesses (e.g. eight for a double-level translation and two for a single-level shadow table translation) caused entries to be put into the DLAT which would not be used in the DL~T

~~7~ 3t7~

and had the undesirable effect of destroying DLAT
entries which had a significant likelihood of future use.

In V~l/370, a user's shadow tables are often purged whenever any page frame in guest main storage is invalidated by the guest. The reason selective page entry invalidation was not done in the shadow tables is hecause it involved excessive software overhead in comparison to purging all shadow page tables, since selective page invalidation would need to search all guest page tables for each page being eliminated from main storage. Therefore, in a VM/370 system, all shadow tables could be purged at a relatively high fre~uency which may result in additional double-level translation overhead if the restoration of many of the purged entries is again requested.

The wide versatility in the services provided by a V~l system make it attractive to system users, who need not be aware of the internal double-level address translations occurring in the operating system soft-ware within the VM system. A significant performance improvement in the IBM VM/370 operating system was obtained in 1974 in many IBM System/370 model machines when the virtual machine assist (V~) microcode was released. Among other things, ~M~ provided a marked reduction in the software overhead in the filling-in of shadow table entries by providing the double-level address translation in microcode in many cases, rather than requiring software intervention. The process in microcode was identical to that of software, except that software interruption handling and status saving and restoring was not required~

PO~-80-005 7 ~'7~i 3~

Also in the prior IB~ ~1/370 system is the special case of a V=R guest talso called the preferred guest, or reserved storage, or selected guest). In the special case of the shadow table bypass mode, only single-level translations are used, even though the preferred guest is operating at the second architec-tural level in the system. This is done by assigning the preferred guest to a predetermined low-address extent of main storage, in which the guest translated real addresses map directly into that extent. In particular, dynamically modified System/360 and System/
370 I/O channel programs can be run by the preferred guest which cannot be used by pageable guests using double-level translation.

However, in the shadow-table bypass mode, the preferred guest addresses are no longer validated by the host program, because the host is not requixed to construct the preferred guest shadow tables. As a consequence, erroneous preferred guest store addresses can damage the host programs. To maintain host integrity under these circumstances (except with respect to I/O), the bypass mode is not used and the host uses shadow tables for the preferred guest to limit all storage accesses by the preferred guest CPU to the preferred guest's extent.

Even when the shadow table bypass mode is used, the host must modify a guest page table to avoid conflict with the host absolute page zero which may be used for the host's prefixed page. With both the bypass mode and the shadow table mode, the translation tables are used to translate reerences to guest real page zero (to give one effect of prefixing), which is needed in V~1/370 on a uniprocessor as well as on a MP
to avoid a conflicting use of absolute page zero by both the host and guest.

P~9-~0-005 ~'7~ 7 The host retains the remaininy part of the main storage above the preferred guest's extent. A section of the retained part of main storage is directly used by the host in which the host operating system can run in either DAT or real addressing mode; and the remaining part is available for use by pageable guests in which the host provides a shadow-table for each user using double-level translations.

In summary, shadow tables protect the host main storage from stray store accesses by any guest CPU
using shadow tables, because the host constructs the shadow tables and thereby assures itself that the corresponding guest is restricted to its host assigned part of main storage. The use of shadow tables for preferred guests at the second level in the system provides host integrity in relation to all pageable guests requiring double-level translations. When the preferred guest at the second level is permitted to use the shadow table bypass mode to avoid the over-head of double-level translations, the host loses integrity in relation to the preferred guest since no shadow tables are then used ~o monitor the guest accesses to main storage, which leaves open the possi-bility that the preferred guest could store into the host's area of main storage and disrupt the operation of the V~1/370 system.

Other prior art which appears to suggest the equivalent of the prior shadow page tables may be the HV associator discussed in column 23 of U.S.A. patent
4,253,145 to R. P. Goldberg. Other limitations described in the claims of patent 4,253,145 are that they only apply to: (1) virtual machines (VMs) which are duplicates of the host machine, and (2) VMs which -9- ~ 7 have a recursive relationship to the host machine. The subject invention does not require either duplicate or recursive relationships between the host and its virtual machines.

A Harvard University Thesis dated October 1972 by R. P. Goldberg entitled "Architectural Principles for Virtual Computer Systems" provides an algorithmic/
mathematical description of a "Hardware Virtualizer (HV) which is generally described without any specific implementation and is to be providable in hardware and firmware (microcode). The thesis and patent 4,253,145 cause each level in a VM system to have its own f-map.
The f-map operates across one VM level (i.e. n+l to n), unlike a shadow table which can operate across two or more levels in a VM system. Thus, Goldberg's VM
system has a large number of f-maps that simultaneously exist in a large multi-level system. Each entry in a HV relates the segment number, page number of a level (n+l) address to a level n address (in the next lower level), and has a validity status bit.

SUMMARY OF THE INVENTION

The advantages and objects of this invention are to provide one or more of the following:

1. To improve the performance of a computer system while it is executing host and guest programs operating under plural levels of architecture in a virtual machine system which may involve plural levels of address translation. For example, the different architectures may use different size addresses, e.g.
one architecture may use 24 bit addresses while another architecture may use 31 bit addresses.

PO9-~0-005 3~7~

2. To provide uniformly good performance of a computer system for different kinds of guest operating systems, e.g. DOS/370, VM/370, MVS/370, etc~ The reduction by this invention is from one software TLB
per process or user (usually involving a large number of software TLBs in the system) to no software TLBs in the V~S system.

3. To modify the one hardware TLB per central processor in the system and use the hardware TLB to eliminate the need for the prior software TLBs. ThUS
in a uniprocessor VM system, only one hardware TLB
and no software TLBs may be required, regardless of the number of VMs, processes or users in the system.
.
4. To improve the performance of a computer system while it is executing a user program using virtual address translation controlled by an operating system running on a second-level emulated or native architecture which is being run under a first-level native architecture on which another operating system also is using virtual address translation, in which both levels share the same hardware machine.
5. To eliminate the need for the shadow segment and page table type of software TLBs, while obtaining an increase in system performance greater than is obtainable with shadow tables when using double-level translations.
6. To eliminate all main storage accesses aEter a double-level translation, in order to directly obtain the result of the double-level translation.

~7~3'7~
7. To provide a hardware TLB which can communicate directly with a double-level translation operation.
8. To provide a novel hardware TLB means which can be shared by guest and host architecture programs which use either single level translations or double-level translation.
9. To provide controls for a hardware TLB for storing a condensed double-level translation into an entry addressable by the guest requested virtual address represented in the entry.
10. To provide a hardware TLB which may contain double-level translation entries and the entire TLB
is not purged whenever any guest page is invalidated or paged out of main storage.
11. To provide a hardware TLB with controls for the selective purging of a particular double-level translation entry repre~enting an invalidated guest page in main storage without affecting other TLB entries.
12. To provide means in a hardware TLB for selec-tively purging all translation entries relating to anintercepted or interrupted guest program without purging other entries in the TLB relating to the host program operation.
13. To provide load inhibit controls with a hardware TLB for inhibiting the loading into the TLB
of some interim single-level translation results occurring within a double-level translation operation, in order to avoid allowing interim entries from des-troying existing valid TLB entries, which could decrease system performance by subsequent retranslations of such PO9-80~005 3~'~

destroyed entries, and which could create a deadlock situation for instructions with at least two storaqe operands when the hardware TLs does not have more than two-way set associativity.
14. To perform any double-level translation of a requested virtual address and to store the resulting main storage address into a hardware TLB entry assigned to the requested virtual address.
15. To select either the guest STO (segment table origin address in main storage) or the host STO for being stored into the TLB entry assigned to a request, in accordance with the request being by a guest program or by the host program.
16. To provide acceleration means with a hardware TLB for enabling a preferred guest using virtual addressing to use the single-level hardware DAT circuits associated with the TLB to perform all translations for the preferred guest.
17~ To provide limit checking means to check preferred guest translated addresses against an upper limit main storage address to determine if each pre-ferred guest request is within the extent assigned to the preferred ~uest to obtain host integrity against preferred guest operation without any system performance degradation, in comparison to the large system perfor-mance degradation caused by the prior art technique of requiring the preferred guest to use shadow tables to protect the host integrity.
18. To provide prefix switching between host and accelerated guest prefix ~alues in the processor's prefi~ register to eliminate the need for the more -13- ~ 7 costly technique of using a guest's page table entry ~o ~lmula~e prefixing for the guest while permanently leaving the host prefix value in the processor's prefix register.

One aspec-t of the invention provides in a data processing system having at least one central processor (CP) and a main storage (MS) for supporting at least one guest virtual machine (VM), the CP including an instruc-tion execution (IE) unit for executing instructions and generating real or virtual address requests for accesses in main storage or a cache, a hardware trans-lation lookaside buffer (TLB) having a plurality of addressable entries, each TLB entry containing a validate field, a STO (segment table origin) field, a virtual address field and an absolu-te address field, each TLB entry also comprising:

an emulation field for indicating whether the entry contains a guest VM or na-tive/host translation, whereby host translations may be maintained in TLB concurrent with guest translations.

BRIEF DESCRIPTION OF T~E DRAWINGS

FIGURES 1, 14 and 15 illustrate the host/guest relationship found within the prior IBM VM/370 system.

FIGURE 2 represents the double translation opera-tion required of a guest virtual address.

FIGURES 3, 4 and 5 illustrate address space and translation rela-tionships found with various GUEST
modes of operation useable within the invention.

FIGURES 6A, 6B and 6C describe the operation of the SIE instruction.

PO9-80-005 ~7~3~ 7 -13a FIGURE 7 represents the interaction between an executing host program calling a guest program in rela-tion to the microcode, hardware and main storage of the central processor executing program. ..

FIGURES 8, 9, 10, 11 and 12 illustrate circuit diagrams representing a preferred emhodiment of the invention.

FIGURE 13 illustrates an MP system in which each central processor (CP) may contain the subject inven-tion.

ros-so-oos 3'~

BACICGROUND OF T~IE II~E~TION

FIGURE 1 illustrates the structure of a virtual machine (~1) system, like the commercial IBM VM/370 operating system in the prior art which is a software system having a microcode assist in various IBM System/
370 Models 158, 168 and 3033. The items illustrated by broken lines in FIGURE 1 represent items removed from V~1/370 by the subject invention, which involves the elimination of the plurality o~ shadow-tables shown in broken lines. The prior VM/370 system provides multi-level operation of architectures of the System/370 and System/360 type and allows different operating systems (such as MVS/370, DOS/370) to operate under the respective architectures provided at the different levels within a single VM/370 system. In FIGURE 1, the lowest level 0 represents the hardware and micro-code CP machine supporting the VM/370 system, and level 0 hardware includes dynamic address translation (DAT) and translation lookaside buffer (TLB) hardware and microcode supporting virtual storage in the operation of the machine under any of the plurality of related architectures including the System/360, System/370 and System/370 extended architectures.
The TLB hardware 1 receives each address translation performed by the DAT hardware 2, and the translation remains in an entry in the TLB as long as the entry space occupied by the translation is not requixed by some new translation. In current commercial large IBM
System/370 systems, the TLB is a set-associative congruence class type of TLB in which each requested address locates a congruence class in the TLB in which the contained entries are associatively tested.
For e~ample each class may contain two separa~e entries to provide two-tay set associativity in the manner well known in the art.

~7~'7~

Level 1 in FIGURE 1 contains the host system control program (SCP) and the host translation tables used by the SCP comprising the host's segment table (ST) 5, and lts related page tables (PTs) 6 along with a plurality o~ host shadow tables 1 through R
shown in broken lines. There is one shadow table for each user 1 through R at level 3.

The shadow tables 1 through R are used by the host program as TLBs for double-level translations of user virtual addresses into main storage addresses.

The host also has translation tables 5 and 6 for translating second-level virtual addresses (guest "real", also host virtual) in the address space into main storage addresses.

The host machine is the real hardware machine at level 0, which has the level 1 architectural inter-face that is used by the host system control program (SCP) which, for examplel may be the control program of a VM/370 ~ype o~ system. The host program directly controls the hardware machine using the level 1 architecture. The real machine at level 0 in VM/370 has tl) a native architecture and ~2) only limited capability to emulate other similar architectures.
The host program uses the native architecture, e.g.
System/370. Any guest may use the native architecture as well as any other available architecture. The particular architecture used by any guest is specified by the host in a software control block. The VM/370 the host mays run with off, in which case the host requested addresses are handled as real addresses, and accordingly are interpreted by the storage con-troller in the hardware machine for accessing absolute ~7~3`~7 addresses in main storage, i.e. the hard~are adds prefixing if the CPU is in an MP (multiprocessor).

Guest virtual machines (VMs) in the system operate at the level 2 architectural interface in FIGURE 1.
Each guest ~l may choose its own operating system (OS) program whi~h then executes at the level 2 architectural interface. Each guest V~l appears to its using programs as a complete system having an assigned set of resources described for each VM in a host's system directory, which contains the name of the ot~mer of each V~l, the amount of "real" storage allocated to each VM in the host address space, and other infor-mation about each VM. Each guest also has its o~n simulated control "hardware", such as control regis-ters, and a PSI~. A guest VM may operate with either - the DAT off or on in relation to its "real" storage.
For example, any level 2 guest VM may use the MVS/370 operating system to support a plurality of users at a ne~t higher level 3 when its users require the guest V~l to run with DAT on in the simulated PSW of the related level 2 guest.

In FIGURE 1, a plurality of the guests 1 through n are pageable guests, and guest (n~l) is a "preferred"
guest.

The pageable guest VM's are each assigned a "real"
storage which is a respective extent in a host virtual storage space. Pageable guest VMs are therefore paged into the system's absolute main storage using the host translation tables 5 and 6 on a demand basis.
Any pageable guest may operate with its PS~ DAT state off or on, depending on whether the pageable guest allows its users to directly manage the respective user assigned "real" storage. For example, a guest ~O9-80-005 7~

operating system like ~S/370 can mana~e its users "real" storage in the guest's "real" storage by providing each user with its own virtual address space which also provides isolation among its users.

FIGURE 14 shows the address mapping within ~/370 of any guest's "real" address space into the host main storage. FIGURE 15 sho~s the address mapping within V~l/370 of any user virtual address space into its related guest's "real" address space, which then maps as shown in FIGURE 1 into main storage.

One user in the VM system may be selected as a "preferred" guest at level 2 by assigning the preferred guest the lowest part of the host's absolute main storage, instead of a part of a host virtual storage space as is done for pageable guests. This requires that the host initially allocate to the preferred guest a predetermined extent of contiguous main storage beginning at absolute address zero, which extent (except for the first page frame) is not thereafter available for use by the host or any other guest. The preferred guest is given the entire responsibility for sub-allocation and management of main storage within this extent and does not use the storage management facilities of the host. The preferred guest may oper-ate with its PSW DAT state either off or on. Whencertain special optional modes of operation are selected, instruction and operand addresses are directly handled as real main storage addresses. In some cases, when the preferred guest's DAT is on, the preferxed guest's addresses are translated by the preferred guest's ST
and PTs directly into the guest's extent in main storage. The guest's tables in some cases are minimally modified by the host program~ When dispatched, the preferred guest's STO is put directly into CRl 8 in t~

FIGURE 1 to access the preferred guest's translation tables, so that the preferred guest's translated addresses are directly provided into the guest's extent in main storage for accessing its data and instructions using only single-level address translation, which is more efficient than the double-level address trans-lation required by level 3 users.

The level 3 architectural interface is controlled by a related level 2 guest VM, and it may be the same as or different from the architectural interface seen by the level 3 users of other level 2 guests. Thus, each set of level 3 users relate to a particular level 2 guest VM which has its o~m SCP which may be the same or different from the SCP being used by any other level 2 guest VMs. For example, guest 1 at level 2 may use the ~S/370 SCP to support a plurality of users 1 through R, each having its own virtual address space at level 3.

In more detail, each level 3 user has its o~m set of translation tables to translate its own level 3 address space into its assigned "real" storage in its related guest's virtual address space at level 2. Also for each level 3 user, the level 1 host has a corres-ponding set of shadow tables (e.g. 3 and 4) in main storage which are used by the host as a TLB for handling the respective level 3 user address requests in the host controlled main storage. For example, if guest 1 at level 2 is using the MVS/370 SCP to support users 1 through R in level 3, then if user 1 is dispatched by MVS, STOl.l (which is the STO for user 1 of guest 1) is fetched by ~S from the guest 1 "real" storage and trans-ferred into the simulated guest 1 CRl, (i.e. Gl (CRl)) to control the address translation for user 1 by means of its own translation tables Ul (ST) and Ul (PT).

Po9-80-005 '7~7 Each entry in Ul (ST) is capable of containing a user's page table origin (PTO) "real" address which locates a user's page table Ul (PT) in the guest's "real" storage. Each Ul (PT) entry may be assigned a page frame "real" address (PFP~) to locate an assigned "page frame" in the guest's "real"
storage, which is part of the host's virtual storage, as previously mentioned.

The translated PFRA output of Ul (PT) is a pageable guest 1 "real'l address which, thus far, involves only a single le~rel of the translation into the guest 1 virtual address space at level 2, which also is a host virtual address space. The resultant user 1 "real" address maps therefore into an assigned page in the host's virtual storage and is received by the host VM as a virtual address request from guest 1, which still must be translated by the host VM to the required main storage location.
This host translation is controlled by the host STO
in CRl 8 with the host DAT on, which addresses the host ST5 and accesses a host PT6 to obtain the next translation level for the user's requested address.

The guest and host translation tables therefore provide two levels of address translation, which involve a total o~ eight accesses to main storage to translate a single MVS user address request into its main storage real address. This address trans-lation process is sho~n in detail in FIGURES 2A and 2B for a user requested virtual address llOA. The following legend is used in FIGURE 2:

~.7~ ~t~7 LEGEND
VA = Virtual Address RA = Real Address SX = Segment Index PX = Page Index D = Byte Displacement in Page G = Guest H = Host HS = Host Shadow Tables STE = Segment Table Entry PTE = Page Table Entry STO = Segment Table Origin PTO = Page Table Origin PFRA = Page Frame Real Address Each circled number in FIGURES 2A and 2B represents a main storage access in the sequence in which they occur during the translation of address 110A. It is seen in FIGURE 2A that the first six main storage accesses are required during the first-level trans-lation (at level 2) to translate the user's requestedaddress 110A into its "real" address 116, which is in the host virtual address space. That is, the user's "real" address 116 (i.e. the guest xeal (GR) address) is then presented to the host which the host considers ~5 a host virtual tHV) address which requires two more main storage accesses by the host second-level trans-lation (at level 1) to obtain the corresponding host real (~IR) address 120 for the requested user data in real main storage.

Therefore, the described double-level translation involves eight main storage accesses for each user address request 110A, which involves a very significant amount of performance overh~ad for the system. This overhead has been reduced by providing the host with Po9-so-005 ~7~ 7 another set of tables (e.g. shadow tables 3 and 4 in FIGURE 1) for each user having its PS~ DAT on (which causes a double-le~el transla-tion).

On the completion of any double-level translation, the resultant page frame real address (PFR~) in main storage is inserted into a correspondin~ entry in the guest's shadow page table 122 in FIGUR~ 2. Shadow ST
121 and PT 122 are also shown in FIGUP~ 1 as the shadow tables 3 and 4 in host level 1 associated with user n in level 3.

The shadow tables 121 and 122 thus represent a translation lookaside buffer for requests by the respective user to obtain a previously double~
translated page address (i.e. user request 110) with only two main storage accesses, i.e. to ST 121 and PT 122.

The double-level translation and shadow table accessing algorithms described in FIGURES 2A and 2s are presently implemented in the prior art by V~
20 microcode in the IBM System/370 Models 135, 158, 168 and 3033. They are described in a prior publication:
IB~I Technical Report TR00.2506 entitled "Virtual Machine Assist Feature Architectural Description" by P. H. Tallman et al dated January 9, 1974. Each shadow PT contains the same number of entries as the associated user's level 2 PT; but the entries in the corresponding shadow PT contain the double-level trans-lation value from level 3 to level 1, while the corres-ponding entries in the user's PT contain only the user's single level trans~ations from level 3 to level 2, which are different values.

PO9-~0-005 ~'7~7~

=22-~ hen any CPU operates with the V~A assist micro-code, a request by a level 3 user first causes the host to access the machine T~B. A miss in the machine TLs causes the respective user shadow tables to ~e accessed in the standard manner (requiring two main storage accesses) to determine if they contain the required translation. Eor example, a request for address 110B in FIGURE 2B by user 1 for an operand first causes the machine TLB to be in~errogated. When the machine TLB does not contain the requested trans-lation, the user 1 shadow table is accessed to deter-mine if this shadow table contains an entry representing the requested page. If such entry is found and it is valid (i.e. its I=0), the entry content Ul (PFRA) is concatenated with the displacement D of the requested user 1 virtual address to generate the main storage address of the user 1 requested data. On the other hand, if the found entry is found invalid, a shadow page table miss signal is generated which initiates the above-described two-level address translation to be e~ecuted (beginning with the same address 110B at the top of FIGURE 2~) by microcode in the manner previously described.

The level 3 interface has been found particularly useful for running users to test program components of an SCP at level 2, wherein a failure in the tested component merely causes one guest to stop running and does not shut down the V~ system which can continue running with other guests operating at level 2.

GENERAL DISCUSSION OF THE INVENTION

FIGURES 3, 4 and 5 illustrate summaries of various translation activities which may be used with this invention. FIGURE 3 represents the operation of a .

3'7~

pageable guest VM at level 2 having a user virtual address space 131 at level 3, which may have a byte size of 224 or 231 dependent on the level 2 architecture selected by the guest VM. Any 4096 byte page in the user address space 131 is mapped to an assigned "real"
page frame in a user "real1' extent assigned to the user within the related guest's virtual address space 132 by the translation from level 3 to level 2 (TRANS 3 to 2).
That is, a user requested address in a user's address space in a guest virtual (GV) address which the guest must translate to a gùest "real" (GR) address.

The guest address space appears to the host as one of the host's virtual address spaces. The host must then recognize the guest's "real" (GR) page frame address as a host virtual (HV) page address, in order to find the corresponding host real address (HR) to ~hich prefixing is added to generate the absolute address in main storage (MS) 133. The MS page frame being used by any pageable guest need not be contiguous in MS.

In FIGURE 3, arrow 130 represents the hardware TLB
translation operation provided by this invention for a user requested address at level 3 to the same MS page frame at level 1 after its double level translation.

The bottom of FIGURE 3 represents the sequence of component address transformations for changing a user requested address at level 3 into a main s~orage absolute address at level 0. The meaning of the illu-strated symbols are: (T) is translation, (P) is pre-fixing.

PO9-~0-005 3~7t~

FIGURE 4 expands the summarized concept in FIGURE
3 to a unique situation in which a plurality of page-able guests operate within a given host virtual address space. Arrows 136 and 137 represent the look-aside translations for the respective user 1 of guest 1 and user 1 of guest 2, respectively. MS
page frames for the guests may be assigned in MS by the host. It is also realized that in any given system using virtual addressing, there may be a large number of guest V~ls, and in turn any single guest may support any number of users for level 3 operation in the system, even though only one user is shown for each guest in FIGURE 4.

FIGURE 5 illustrates a sumrnary of the translation operations for a preferred guest in a VM system which also has pageable guests. The single preferred guest is sho~n as a preferred guest virtual address space for which the guest's DAT is on. With virtual addressing, the preferred guest's address space may be 224 or 231 depending on the architecture selected by the preferred guest VM. If the guest's DAT is off, the guests level 2 address space will be "real" and will map directly into the preferred guest's main storage extent, which is assigned main storage addresses from 0 to K when the preferred guest is initialized by the host. The host is thereafter restricted to using the page frames in main storage having addresses from K to M, ln which the host program and data are contained, and from which the host can assign page frames for use by pageable guests on demand. Thus, the host SCP may not there-after allocate page frames within the preferred guest'sextent from 0 to K. The preferred guest assurnes full responsibility for main storage management within the guest's extent. When the preferred guest's DAT is off, the guest's address requests can be directly accessed in the guest's extent 0 to K, because the guest's Po9-80-005 74.-;9 ~3 ~ 7 addresses are not necessarily then translated. When the preferred guest's DAT is on, it provides its own single level DAT ST and PT in main storage (not shown in FIGURE 1, but similar to the host ST and PT) which can be used to translate the preferred guest requested addresses into main storage real addresses in the preferred guest's extent. The preferred guest has a performance advantage over pageable guests due to each preferred guest address translation requiring only single level translation as compared with page-able guest double-level address translation requiring eight main storage accesses.

FIGURE 13 illustrates an example of a central electronic complex which may use the subject invention.
Although FIGURE 13 shows multiprocessor (MP) hardware with four central processors (CPs), the invention will operate with a minimum configuration of a single CP (i.e. a uniprocessor) connected to one BSM (basic storage module), and an I/O EXDC (external data controller) having the I/O channels of the system.

FIGURE 7 illustrates a central processor (CP) containing the invention, which may be a UP or may be any one of the CPs in the MP shown in FIGURE 13. The CP in FIGURE 7 includes instruction unit (IE) hardware 20 which may be of conventional type except for the addition of SIE instruction execution means which is ~escribed later herein.

A programming system of the type represented by the IBM VM/370 control program (but modified to include the SIE instruction) is program loaded into the system to support a host which accommodates a plurality of guest virtual machines (VMs). The CP is also loaded with microcode including VMA type microcode for per~
forming double-level address translation.

3~7 The host maclline operation is initiated by any instruction stream dispatched by the host control program in the conventional manner. In FIGURE 7, a host program is illustrated by the host instruction stream 21. A guest virtual machine operation is initiated by the host instruction stream 21 executing a start interpretative execution (SIE) instruction 22, which fetches fields from main storage in a state description (SD) control block located by the ~2, D2 fields in instruction 22. The SIE instruction is exe-cuted by the I~ in the CP using its B2, D2 address as a real main storage address to fetch the various fields of SD23 from main storage (MS) and load them into special triggers and circuits provided in the CP, which thereafter cause the CP to operate as the guest VM.
The SD thus describes the architecture for the guest V~l which will be seen by the guest instruction stream 33. The CP thereafter emulates the guest machine defined in the fetched state description as long as the guest program is executing. Thus, the guest instruction stream is initiated at a nested level under the host instruction stream by operation of the SIE instruction which, when executed, changes the CP to emulate the guest machine which then executes as the guest VM
defined by the state description in the SIE instruction.

Among the CP circuits and triggers loaded by SD
fields are a guest/native mode trigger 26 which is set to guest mode, a preferred/pageable guest mode trigger 27, an accelerated/non-accelerated preferred guest mode trigger 28 and an emulated/native architecture trigger 29. Other hardware loaded by SD fields includes prefix and extent limit registers, as well as control registers (CRs), program status word (PSW) registers, STO registers, general registers (GRs), etc. which are in the CP in FIGURE 7 but are not all sho~ in detail.

3~7~7 The execution of the SIE instruction also causes the current state of the host machine to be saved by the IE.

FIGURES 6A, 6B and 6C are a flow diagram providing a detailed self-explanatory description of the micro-code cution of the SIE instruction.

Once initiated by the SIE instruction 22, the guest program continues its execution in the guest VM with the guest instruction stream 33 until an interception (exit) event or an interruption event is detected by IE
detector 34 or 35, respectively. An interruption event is, for example r an I/O or external interruption requiring the host programming services. An inter-ception event is, for example, the execution of any one of a number of instructions by the guest which require host programming services. A mandatory interception event occurs when the guest attempts execution of certain pxivileged instructions, while other privileged instructions may be optionally flagged by the interception mask received from the SD provided by SIE instruction 22 to cause an interception whenever any flagged instruction occurs during the execution of the guest instruction stream. On the occurrence of an interception or interruption event, the guest instruction stream 33 and the guest VM terminate, and the CP is restored to its host state saved when the SIE instruction was executed. That is the CP hardware is loaded from the saved information put into its LS
(Local Store) and system area in MS when SIE was last executed. Then the host stream 21 again begins -to execute the host instruction stream 21, starting with the host instruction following SIE instruction 22.

Pos-so-0o5 The following TABLE 1 represents the various native and guest CP modes which can exist in the preferred em~odiment described herein. These CP modes are deter-mined by combinatorial settings of the mode trig~ers 26, 27 and 28 in FIGURE 7, as follows:

CP NATIVE/GUEST MODES

MODE TRIGGER STATES

CENTRAL PROCESSOR Guest Mode Pref. Guest Accel.Pref.
MODE T Mode T Guest Mode T
Native ~lode Off Off ff Guest Pageable Mode On Off Off Guest Non-Accelerated Preferred ModeOn On Off Guest Accelerated Preferred Mode On On On The guest mode setting of trigger 26 occurs on execution of a SIE instruction and indicates that the CP is set to emulate the guest architecture and state defined in the SD of that SIE instruction. The native mode setting of trigger 26 occurs when the host is dispatched and when a guest is terminated and indicates the executing program only sees the host defined archi-tecture of the CP.

Trigger 27 is effective when trigger 26 is set toguest mode. Trigger 27 indicates whether the CP will operate as preferred guest V~1 or a pageable guest VM.

Trigger 28 is operative if trigger 27 indicates a preferred guest mode. Then trigger 28 indicates whether the preferred guest VM will operate in an Pos-so-005 ~7~3~7 accelerated mode. The accelerated mode enables the preferred guest to use the hardware dynamic address translation (DAT) circuits in the scE to perform the single level translations required by a preferred guest. If non-accelerated mode is indicated by trigger 28, then preferred or pageable guest trans-lations are not done by the BCE hardware but instead are done by IE microcode, with the microcode trans-lation results being loaded into the hardware TLB 132 in BCE 31 shown in FIGURE 10.

An emulated/native architecture type trigger 29 indicates whether the CP architecture used by the current guest is the native architecture of the machine or whether it is an emulated architecture twhich is different from the native architecture. In the des-cribed embodiment, the native architecture uses 31 bit addresses with DAT on or off in instructions of the System/370 type. The emulated architecture uses 24 bit addresses with DAT on or off and may be System/370 or System/360. With DAT on, the native architecture can therefore support plural user address spaces each having 231 bytes, while the emulated architecture can support user address spaces each having 224 bytes.

All of the triggers 26, 27 and 28 are reset by a reset signal provided from IE 20 upon the occur-rence of an interception or an interruption terminating the guest mode.

FIGURE 8 illustrates in detail the access control circuits 31 shown generally in FIGURE 7 and found within the IE and BCE. The access mode controls in FIGURE 8 receive the outputs of the IE mode triggers 26, 28 and 29 as well as micro-order command outputs on bus 25 from FIGURE 7. In FIGURE 8, logic circuits Ç`3~

~ 30-are provided between the received signals and inputs to a plurality of storage access control triggers 41, 42, 43 and 44.

The settings of triggers 41, 42, 43 and 44 control the manner of operation of the translation circuits in FIGURE 9 and the DLAT circuits in FIGURE 10 in response to each CP storage request.

These triggers are reset by the OR of any of the seven micro-order signals shown in TABLE 2, and a trigger will be set on if it receives a simultaneous set and reset. Non-accelerated guest access trigger 41 is set by AND gate 53 when it receives a guest mode input from trigger 26, a non-accel~rated guest mode input from trigger 28 via an inverter, and the output of OR circuit 51 which signals a guest request when receiving any of the microorder guest request signals Sl, S2 or S3, which are defined in TABLE 2 below:

REQUEST TYPE MICROCODE ORDER SIGNALS

MIRCOCODE "
QRDER REQUEST TYPE
Sl Virtual (Host or Guest) S2 Real (Host or Guest) 25 S3 Logical (Host or Guest) S4 Host Logical S5 Host Real S6 Host Absolute S7 Native/Host System Area PO9-8~-005 -31- ~ 3~7t~
In FIGURE 8, the accelerated preferred guest access trigger 42 i5 set when an accelerated preferred guest signal is provided by trigger 28 while OR circwit 51 is receiving access mode request microorder signal Sl, S2 or S3. An inhibit DLAT load (IDL) signal is provided by an AND circuit 54 when it receives a guest mode signal from trigger 26 and a non-accelerated guest mode signal from the inverted output of trigger 28.
~hile the IDL signal is on, the DLAT cannot be stored into by the hardware DAT logic.

TABLE 3 below indicates the various storage access types for IE requests to the BCE, as determined by the settings of the guest trigger 26, the accelerated pre-ferred guest mode trigger 28, and the inhibit DLAT
load signal (IDL) on line 54A.

CP STORAGE ACCESS TYPES

Non-Accel. Accel.Pref. Inhibit Guest Acc. Guest DLAT Load 20 ACCESS TYPE T Access T Signal Native Access Off Off Off Host DLAT Access (In Guest ~lode)Off Off On Non-Accel. Guest 25 Access On Off On Accelerated Pre-ferred Guest Access Off On Off Also in FIGURE 8, native architecture access trigger 43 indicates whether or not the requested access is using the native or an emulated architecture. Trigger 43 is set by the output of OR circuit 67 when it receives an input from either an AND circuit 57 or 61. AND circuit PO9-80-~05 57 is activated by a signal from OR circuit 51 and a native architectural type signal from trigger 29 through an inverter. AND circuit 61 is activated by a signal from the output of an OR circuit 52 and the output of an OR circuit 58 providing either a native architecture type signal from trigger 29 via an inverter or a signal from guest mode trigger 26.

Virtual access trigger 44 is set by the output of an OR circuit 68 when a virtual address is requested by any virtual request microorder signal to OR circuit 51 or 52, as defined in TABLE 3 above in which a logical request provides a virtual address if the DAT bit (PSW 5) is on in the PSW of the requesting machine, native or guest.

FIGURE 12 shows special command circuits for controlling BCE operations relating to guest or host requests. These special commands control the loading of BCE registers, the writing into a DLAT entry, and the purging of the DLAT entries. In FIGURE 11, the microcode controls 20A in IE 20 issue each command as a set of micro-order signals on bus 25 which includes lines 173 and 174 to a special access trigger 83 and line 175 to a speeial aceess micro-order register 82 in FIGURE
12. Miero-order signals are provided on lines 175A-175H
at the same time that special access trigger 83 is set by a signal from IE 20 on line 174. The output of trigger 83 activates a decoder 84 in the BCE to decode the micro-order signals set in register 82 to aetivate one of a plurality of output lines from deeoder 84 aecording to the particular command being signalled.
Outputs 91-98 are pertinent to the subject invention.
The other outputs provide other functions which are not pertinent and therefore ~ill not be diseussed.

PO9-~0-005 ~ 'i The decoded command signal on line 95 or 96 contrc)ls the loading of the host STO or guest STO
into the BCE circuits in FIGURE 9 by activating either gate 121 or 123 to transfer the host STO or the guest STO from an IE storage ~U5 (SDB) into either register 122 or 124. A signal on lines 97 loads the prefix and limit registers 107 and 102 in FIGURE 9. It is only after the initializing of the STO and prefix registers that the access control circuits in FIGURE
8 through 10 can operate to control the selection and read out of an entry in the DLAT 132, or control the writing of an entry in DLAT 132, or inhibit the writing in the DL~T of certain types of entries which would degrade DLAT operation. The decoded 15 commands on lines 93 and 94 in FIGURE 12 control different manners of purging entries in the DLAT.

For a DL~T operation, either the host or guest STO must first be selected in FIGURE 8. STO selec-tion is controlled by the output state of the non-accelerated guest access (NAGA) trigger ~1 in FIGURE8 which provides its NAGA output to FIGURE 9 to activate either gate 128 or 127 to select either the host STO or the guest STO for transmission on lines 129A to provide the particular STO used for the selec-tive readout or write-in of an entry in DLAT 132.

In FIGURE 8, a G (guest) flag signal is provided on lines 72A and 72B to identify a DLAT entry as representing a guest or host translation. A host entry is indicated when G is a zero bit, and a guest entry is indicated when G is a one bit. The G flag signal is generated by the output of OR circuit 73 when it receives a signal from trigger 41 indicating whether the request is for a non-accelerated guest, or a signal from an AND gate 77 indicatin~ a virtual access request by an accelerated preferred guest due to signals from triggers 44 and 42.

An R flag signal is provided on lines 45A and 45B
to identify if a guest request (i. e. in which G=l) provides a virtual address or real address. The R
flag is a zero bit if the guest requested a virtual address, and is a one bit if the guest requested is a real address. The R flag is the inverted output 45 of virtual access trigger 44.

Also, FIGURE 8 provides signals on lines 42A, 42B, 43A and 44A to control the DAT operations done in FIGURE 9, which is done for accelerated preferred guest requests as well as for native/host requests, all of which require a one-level translation.

The inhibit DLAT load (IDL) signal on line 54A
to the DLAT write controls in FIGU~E 10 controls the writing into a selected DLAT entry of an address trans-lation for the current IE request. The G and R flag signals are provided on lines 72 and 45 to control the DLAT operations for guest requests. Thus, the G flag indicates if the current request is from a guest or the host, while the R flag indicates if a current guest request is using a virtual or real address.

The method by which this invention switches the guest prefix values into the CP prefix register is novel with this invention, and eliminates the need to use a guest's page table to simulate prefi~ing of the guest's real page zero.

PO9-~0-005 63~7t7 Thus in the embodiment, when the SIE instruction is executed, its microcode: (1) in FIGUR~ 6~ stores the host prefix into the CP local store from prefix register 107 in FIGURE 9; (2) in FIGURE 6B determines if the accelerated mode is to be used, then sets accelerated guest mode trigger 28 in FIGURE 7, and loads the preferred guest prefix value obtained from SD into prefix register 107.

~hen the SIE instruction execution completes due to interception or interruption, the host prefix is moved from local store into prefix register 107.

A guest ~IS limit value is loaded from the SD
into limit check register 102 at the same time that the guest prefix is being loaded from the SD into prefix register 107, which values are simultaneously transferred through different fields in D register 81.

While the accelerated preferred guest mode exists, both limit checking and prefixing of the guest's translated addresses are accomplished at no cost in CP performance, as occurred in prior VM/370 systems. Limit checking determines if each request is within the MS extent provided for the preferred guest by the SD of the last loaded SIE instruction.

Host program storage references are not limit checked. ~licrocode originated re~erences to host main storage are not subject to host prefixing because they are references to the SD which should not be in any prefix or reverse prefix area in main storage.

3 d A prefix/limit command on line 97 from FIGURE 12 to FIGURE 9 causes the transfer of an upper limit (UL) value and a prefix value from different fields in a word sent by the IE through a D register 81 into upper limit register 102 and prefix re~ister 107 under the control of acceleration mode trigger 28.

One active preferred guest may exist as shown in FIGU~E 5 r and the guest's lower limit is a zero displace-ment value and therefore is not checked in F~GURE 90 FIGURE 9 also illustrates a conventional type of dynamic address translation (DAT) adder 113 and DAT
sequencing controls 117, which provide a one-level translation for a requested virtual address involving two main storage references of the segment table addressed by the STO on lines 122A to the TR register 114, and the page table addressed from the PTO
accessed in the segment table entry which is provided from main storage bus out ~SBO) to the TR register 114. The VAS register 112 receives the SX and PX
components of the virtual address during the trans-lation, and each resultant address from DAT adder 113 is provided into D register 81 and transferred to main storage or accessing. The DAT sequencing controls 117 control the DAT translation operations under control of its input lines, such as line 42B which controls a translation for an accelerated preferred guest, line 43A which controls a native/hosk trans-lation, line 44A which indicates if the request is for a virtual or real address, and line 98 which is a special command line activated by the IE when it executes a load real address (LRA) function which causes the DAT adder 113 to translate the IE supplied virtual address with the resultant translated output Pos-so-0o5 ~7~ 7~

of DAT adder 113 being sent back to the IE instead of to main storage so that no actual main storage access results and the translation is not entered into DLAT
132.

Host requested real addresses not requiring translation (i.e. host DAT off~ are not entered into DLAT 132. Also, accelerated preferred guest real addresses not requiring translation (i.e. preferred guest DAT off) are also not entered into DLAT 132.
Thus, the only host/native or guest requests requiring translation (indicated by virtual access trigger 44) are put into DLAT 132. Requested real addresses not requiring translation are prefixed (if necessary) and sent to main storage for accessing, and are not sent to DLAT 132.

FIGVRE 10 illustrates the DLAT (Dynamic lookaside translation) buffer 132 which contains translations that have recently been done for either a guest or a host request. DLAT 132 is presumed to be set-associative in this specification in the conventionalmanner of set-asso~iative DLAT's, e.g. two-way set associative DLATIs are used in current commercial IBM
large systems such as the 3033 central electronic complex. However, novelty exists in DLAT 132, which differs from prior DLAT's by con~aining in each entry a guest tG) bit and a real (R) bit, in addition to the prior used fields which include an invalidity (I) bit (which sometimes is called a validity (V) bit), an STO field, a virtual address (VA) field and an absolute address (AA) field. A row in DLAT array 132 is called a set-associative congruence class, because it is presumed to have plural entries (even though only one entry is shown in FIGURE 10 per congruence class).

1~76~7~

Each class is selected by the output of an address decoder 136 ~hen it receives the current IE requested virtual address provided from a DLAT address reglster 137.

A set of DLAT compare circuits 151, 152, 153 and 154 is provided for each entry in a set-associative class, (one set is shown in FIGURE 10) to determine if any of the entries in the selected congruence class represents the requested address. (The prior art teaches how plural sets of compare circuits can be used for set-associative DLAT ' s . ) If the compare circuits find that the requested address is represented by an entry in the selected congruence class, a DLAT
hit signal is provided on line 165. If the requested address is not represented in an entry in the selected class, then a DLAT miss signal is generated on either line 163A or 164A. The DLAT miss signal on line 163A
can occur for a native/host request or an accelerated preferred guest request, any of which require a single level translation by the DAT hardware in the BCE. A
DLAT miss signal on line 164A can occur for a non-accelerated guest access (NAGA) request.

A miss signal is generated when less than all of compare circuits 151, 152, 153 and 154 provides a com-pare-equal output (except for a common segment indi-cation which does not require STO equality). AND gates 156 and 158 combine all equal outputs from gates 151, 152, 153 and 154, (including a forced equality from an OR circuit 157 if a common segment is involved) to generate a hit or miss output.

In more detail, compare circuit 151 compares the G bit outputted by a selected DLAT entry and the current request's G flag provided on line 72A. Compare ~09-80-005 63~

circuit 152 compares the R bit outputted by the selected DL~T entry and the current request's P~ flag being pro-vided on line 45~. STO comparator 153 compares the STO in the selected DLAT entry wi-th the STO being provided for the current request from the selected STO
register 122 or 124 in FIGURE 9, and virtual address comparator 154 compares the virtual address bi~s in the DLAT entry and corresponding vir~ual address bits of the current re~uest being provided on line 81A
from FIGVRE 9. Thus, if the outputs of all comparators 151, 152, 153 and 154 indicate an equal condition, a "hit" exists for the current re~uest, which is then signalled on line 165 and accessed in the DLAT entry having the equality by conventional circuits not shown in this specification. Conversely, a DLAT
miss occurs as an output from inverter 162, which is selected by either AND gate 163 or 164, depending upon whether the request was made by a non-accelerated guest request, or by a host/native or accelerated guest request.

If a request misses in the DLAT, then the trans-- lation of the request must be written into an entry in the selected DLAT congruence class whereby the entry is selected in the conventional manner by an LRU circuit (not shown). The DLAT entry is written by DLAT write control 131 in FIGURE lOo It is acti-vated by the output of an OR circuit 133 while any of the three illustrated inputs is activated. One acti-vating input is from an AND gate 134 which is activated by a DAT write signal on line 117A and no IDL signal from inverter 135 receiving the IDL line 54A. When write control 131 is activated, it writes into the selected D~AT entry the one or zero states existiny on the invalid bit line 143, the G flag line 72B, the .. _ _ , .. .. ..

i 3'~

R flag line 45B, the STO lines 129~, the virtual address lines lllB, and the absolute address lines 108~ from FIGURES 8 and 9. The I bit is set valid in the entry being written. ~n intermediate state may also exist on any of these lines which is not the one or zero state and which does not change the corresponding field in the entry being written.

A miss signalled on line 163A is provided to the DAT sequencing controls 117 in FIGURE 9 to activate a DAT operation by adder 113 which in~ernally operates in the manner of DAT cixcuits in current machines such as the IBM 3033 CEC. The difference from the prior art is in the controls provided to initiate a DAT operation and to select the STO used for the trans-lation.

A miss signalled on line 164A is provided to theIE to initiate a microcode translation of the guest address, which may be either a single-level or double-level translation, as required by the type of request.

The IE unit will perform the guest segment table lookup by adding the guest STO to the SX portion of the virtual address and then adding the host offset value to obtain a host virtual address. In FIGURE 11, this resultant host virtual address is provided by the IE on line 177 via OR circuit 172 to D-Reg 81 in FIGURE 9 with triggers 41 and 42 reset. This address is then gated into VAS register 111 for translation procedures if a DLAT miss should occur. The host STO
register 122 content is put into TR register 11~ and controls 117 activate gate 112 to output the SX from register 111, so that the ST entry address is generated by adder 113 and put into the D register 81, from which it is sent to main storage (MS) as an MS request.

~7~3~7~

The ST entry returns on ~S SBO ~storage bus out) and is put into TR register 114. Controls 117 then enable gate 112 to output the PX field in VAS register 111, so that adder 113 then generates the PT entry address which is sent to D register 81 and requested in ~S. The PFRA address is then accessed in MS and returned on MS SBO into TR register 114, from which it is passed through adder 113 along with the page o~fset bits from the VAS register into D register 81.

Since line 54A is active, the DLAT write will be inhibited. The address in the D-reg is sent to main storage (MS) as an MS request. The data from this access, which is the guest segment table entry, is returned to the IE unit. A similar type of procedure is performed to obtain the guest page table entry.
The results of that access are then concatenated with the guest page offset added to the host virtual offset and presented to the BCE on line 177. Instead of a Fetch operation, a signal on a LRA line 98 in FIGURE
12 is issued which controls the DAT sequencing controls 117 to return the page table entry to the IE unit.

The IE unit will issue a special command called the DLAT A command on line 91 and pass the guest virtual address on line 177. The DLAT A command will lock this virtual address into the VAS register 111. The IE
unit will then issue a special command called the DLAT B command on line 92 and pass the host absolute address on line 177~ Also, trigger 41 is set on along with the other mode triggers (43-44) that describe the type of guest. The DLAT B command will activate the DL~T write circuit 133 to write the selected DLAT
entry in accordance with the signals on lines 143, 72B, 45B, 129B, lllB and 108B to complete a BCE
single-level translation entry in DLAT 132.

~7~3~7 The hardware DAT operation is used by both native/
host requests and by accelerated preferred guest requests.
In either case, the single-level translation has the same fast speed of operation.

The IE determines whether a guest request trans-lation will be accelerated or not. All native/host translations are accelerated, i.e. done by the BCE
hardware.

A non-accelerated guest request is determined if the guest request cannot be translated by operation of the BCE hardware, i.e. DAT sequencing controls 117 and DAT adder 113. If a non-accelerated guest request is to be made, the IE 20 sends a set signal to trigger 28, and the DAT adder 113 and sequencing controls 117 will lS not operate to perform the translation and the trans-lation will be done by IE microcode. Non-accelerated guest requests include preferred guest requests which for some reason cannot be handled in the ~CE, e.g. the limit register is the wrong size for checking the request upper limit, or the virtual address size is not compatible with the DAT adder si~e such as the adder input has improper alignment of the SX and PX
positions in a requested virtual address, etc. All preferred guest translations are one-level translations involving two main storage accesses, regardless of whether the preferred guest request is accelerated or not, but the BCE hardware operates faster than IE
microcode performing the same translation and hence the BCE "accelerates" the translation.

Pageable guest requests are never accelerated in this embodiment, i.e. do not use the BCE DAT adder 113 and DAT sequencing controls 117, but are done by IE microcode. A pageable guest may therefore initiate 3'~'7 a double translation operation by VM~ type microcode in the IE microcode execution hardware 20B in FIGURE 11.
The microcode is provided ~rom the microcode storage 20A, which may operate in the conventional manner to execute a double translation as previously explained for loading a shadow-table in regard to FI~URE 2 herein.

However, with this invention, no shado~l-table is used. During the execution of any translation by IE
microcode, the novel IDL signal is provided on line 10 54A to inverter 135 in FIGURE 10 to deactivate -the DLAT write controls 131 during the existence of the non-accelerated guest mode. ~hen the translation is completed by the microcode, the resulting absolute address is provided from IE 20 in FIGURE 11 on absolute 15 address bus 179 via OR circuit 172 and bus 177 to D-reg 81 in FIGURE 9, from which it is provided on lines 108A to the DLAT controls 131 in FIGURE 10 at the same time that the valid bit, G. R. STO and VA
inputs are active to write controls 131 for the translated request. At the same time that IE 20 provides the ,ranslated absolute address, IE 20 also provides special micro-order signals on lines 175 to microorder register 82 in FIGURE 12, which are decoded to activate the DLAT B signal on line 92 to OR circuit 25 133 in FIGURE 10 which activates the DLAT write controls 131 to write the translation into the selected DLAT entry.

This invention enables the IE to selectively purge guest entries from DLAT 132, without affecting any host entries in the DLAT. A guest purge, for example, may be done when t~e CP state is switched from guest state back to host state when a guest program ends execution due to an interception or ~ ~7 interruption. This allows the host program, which then begins e~ecution, to use the available TLB
entries in the DLAT without needing to invalidate any host entry, which potentially increases the system operating efficiency.

Thus, when a guest pro~ram loses the CPU
resource, IE 20 in FIGURE 11 can automatically issue a guest DLAT purge command on lines 175 when-ever it issues a reset signal to guest mode trigger 26 in FIGURE 7 to cause a guest DLAT purge signal on line ~4 in FIGURE 12.

The DLAT purge controls 140 in FIGURE 10 receive a signal on line 94 to condition a DLAT cycle control 138 which cycles through all of the class addresses possible in DLAT 132, causing each DLAT entry to have its I and G bits outputted from DLAT array 132 to an AND gate 141 which is being conditioned by the signal on line 94. Thus, AND circuit 141 is enabled by each valid guest DLAT entry (i.e. I=l and G=l) and provides its output through an OR circuit 142 and inverter 144 to the invalid bit input line 143. The output of OR circuit 142 is provided also to OR
circuit 133 to activate the DLAT write controls 131 and cause the entry currently being read out to have its invalid bit set to the invalid state of zero. ~hen the class cycling is completed by the DLAT cycle con-trols 138, all guest DLAT entries are invalidated, but the state of each host entry in the DLAT is not affected because its G bit is in zero state which prevents gate 141 from being enabled by any hos-t entry.

Of course the conventional type of purge TLB
instruction may be done in DLAT 132 by a special command which activates the purge line 93 from decoder 3~

84 in FIGURE 12 to FIGURE 10 which causes all entries to be invalidated by having line 93 provide an input to DLAT cycle control 138 vla OR circuit 146, and to OR circuit 142 which sets invalid every DLAT entry as it is being cycled, ignoring the G and R bits in the DLAT entries.

Hence, a limit check operation occurs in the embodiment whenever an accelerated preferred guest ma~es an access request. The access request address is provided by the IE to D-reg 81. If the preferred guest has DAT off, D-reg 81 receives a real address which is immediately limit checked. If the preferred guest has DAT on, the limit checking is not done until the real address is available, which is after the guest virtual address is translated, which is provided from the D-reg 81 to the limit check hardware. The timing of the limit checking is done by inputs to AND
gate 104. Line 44C times when a request is received with DAT off. Line 117A times when the real address 20 is available from a translation. Lines 44C and 117A
are inputs to an OR circuit 105 which output times the operation of IVA gate 104. Comparator 103 inputs to gate 104 when the specified upper limit is exceeded by the request. The APGA line 43A conditions the IVA
gate 104 so that only accelerated preferred guests can enable it. The input from the APGA line 43A may ~e eliminated so that all preferred guest requests are limit checked if all the limit hardware supports such checking. Thus, if comparator 103 finds that the address D is greater than UL, an invalid address is signalled from an AND gate 104 for a preferred ~uest signal from line 44C indicating that a real (or absolute) access is being tested in comparator 103, and from line 43A indicating an accelerated preferred guest address is being tested.

3~

An invalid address (IVA) output from gate 104 causes an address speciication exception in the - conventional manner to the central processor so that it can generate a program interrupt in the standard manner.

While the invention has been particularly shown and described with references to preferred embodiments thereof, it will be understood by those skilled in the art that the foregoing and other changes in form and details may be made therein without departing from the spirit and scope of the invention.

Claims (25)

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:
1. In a data processing system having at least one central processor (CP) and a main storage (MS) for supporting at least one guest virtual machine (VM), the CP including an instruction execution (IE) unit for executing instructions and generating real or virtual address requests for accesses in main storage or a cache, a hardware translation lookaside buffer (TLB) having a plurality of addressable entries, each TLB entry containing a validate field, a STO (segment table origin) field, a virtual address field and an absolute address field, each TLB entry also comprising:

an emulation field for indicating whether the entry contains a guest VM or native/host translation, whereby host translations may be maintained in TLB concurrent with guest translations.
2. In a data processing system as defined in Claim 1, the CP further comprising:

a guest mode bistable device for being set by the IE executing a SIE (start interpretive execution) instruction in a program of the host machine to indicate the CP has a guest emulation state.
3. In a data processing system as defined in Claim 2, the CP further comprising:

access control means for directly loading an entry in a TLB representing the initial translation for any address requested by a guest program initiated by the SIE instruction, whereby the TLB may be loaded by either hardware or microcode.
4. In a data processing system as defined in Claim 3, the CP further comprising:

means for registering a host STO (segment table origin absolute address) and a guest STO, STO selection means for selecting either the host STO or the guest STO registering means.
5. In a data processing system as defined in Claim 4, the CP further comprising:

IE request means for signalling to the TLB
each IE storage request including a guest/host indicating signal for indication whether each request is by a guest program or by a native/
host program, compare circuit control means for receiving a TLB output for each IE storage request including an entry STO provided by an outputted TLB entry and comparing the entry STO with the guest STO or host STO selected by the STO selection means, whereby a TLB miss signal is generated by the compare circuit control means if the compared STOs are not equal.
6. In a data processing system as defined in Claim 5, the CP further comprising:

the TLB compare circuit control means also receiving and comparing the guest/host indicating signal with a corresponding position in the TLB
output to signal a miss in the TLB if a non-equal comparison occurs.
7. In a data processing system as defined in Claim 6, the CP further comprising:

a real/virtual address field for indicating whether the entry contains a guest VM real address or a guest VM virtual address,, whereby the real/virtual address field in combination with the emulation field can indicate whether the entry contains a guest VM real address translation or a guest VM virtual address trans-lation, and host address translation may be provided for both a guest real address request and a guest virtual address request, the IE request control means also signalling to the TLB with each guest storage request a real/
virtual indicating signal, the TLB compare circuit control means also receiving and comparing the real/virtual indicating signal with a corresponding position in the TLB
output to signal a miss in the TLB if a non-equal comparison occurs.
8. In a data processing system as defined in Claim 7, the CP further comprising:

a preferred guest type indicating bistable device for being set by the IE executing a SIE
instruction to indicate a preferred guest which directly addresses part of the host main storage space without any host address translation, hardware adder translation means being acti-vated by a miss signal when the preferred guest bistable device is set and the preferred guest provides a virtual address request.
9. In a data processing system as defined in Claim 7, the CP further comprising:

a pageable guest type indicating bistable device for being set by the IE executing a SIE
instruction to indicate a pageable guest which directly addresses part of the host virtual address space which requires host address trans-lation to address main storage.
10. In a data processing system as defined in Claim 7, the CP further comprising:

a non-accelerated guest indicating bistable device being set by the IE executing a SIE
instruction when the IE determines that storage requests by the guest cannot be translated using a hardware translation mechanism in the CP but instead must use microcode translation means in the IE.
11. In a data processing system as defined in Claim 10, the CP further comprising:

microcode translation means in the IE, a non-accelerated guest miss bistable device being set by the guest TLB miss signal when the non-accelerated guest bistable device is set, the microcode translation means being actuated to execute a translation of the IE requested address causing the guest TLB miss signal when the non-accelerated guest bistable device is set, whereby a non-accelerated guest requested address is translated by the IE microcode trans-lation means instead of the hardware adder trans-lation means.
12. In a data processing system as defined in Claim 9, the CP further comprising:

TLB write means for writing a translation into a selected TLB entry, TLB inhibiting means for inhibiting the TLB
write means from writing a translation into the currently selected TLB entry, IE command means for signalling the TLB write means to write a resulting translation into the selected TLB entry, whereby the result of a two level translation of a guest requested address can be entered into the selected entry while intermediate translations which are components of the two level translation are not entered into the TLB and therefore cannot displace any potentially good TLB entry.
13. In a data processing system as defined in Claim 7, the CP further comprising:

guest purge control means being provided with the hardware TLB, the IE command means signalling a guest purge command when guest entries are to be selectively removed from the TLB, selective invalidation means connected to the TLB for invalidating each entry in the TLB currently having a guest indicating signal.
14. In a data processing system as defined in Claim 13, the CP further comprising:

the selective invalidation means further including, TLB cycling means for cycling the outputting of the classes of entries in the TLB, whereby the guest purge control means invali-dates the guest TLB entries without affecting any other entries in the TLB.
15. In a data processing system as defined in Claim 14, the CP further comprising:

means in the IE for requesting a fetch of a state descriptor (SD) at a real address in MS
designated by the SIE instruction being executed by the IE, loading means in the CP for setting the states of the guest mode bistable device, a preferred guest mode bistable device, control registers, general registers, local storage, and program status word registers according to information in predetermined fields in the SD.
16. In A data processing system as defined in Claim 15, the loading means further comprising:

means for setting an architecture type bistable device according to a field in the state descriptor to determine the architecture for a guest virtual machine being prepared for execution on the host machine, whereby the guest virtual machine may operate by a different architecture from the host machine.
17. In a data processing system as defined in Claim 16, the CP further comprising:

interruption detection means for resetting the guest mode bistable device to native mode state, whereby the guest program ceases execution and the host program again begins execution.
18. In a data processing system as defined in Claim 16, the CP further comprising:

interception detection means for resetting the guest mode bistable device to native mode state, whereby the guest program ceases execution and the host program again beings execution.
19. In a data processing system as defined in Claim 8, the CP further comprising:

the STO selection means selecting the host STO registering means in response to the preferred guest type indicating bistable device being set to indicate a preferred guest is currently executing in the CP, means for connecting the host STO registering means to the hardware adder translation means for translating preferred guest storage requests, whereby the preferred guest uses the host STO
registering means.
20. In a data processing system as defined in Claim 7, the CP further comprising:

SIE instruction execution means invoking microcode for determining if accelerated guest conditions exist for permitting parallel use of hardware facilities in the CP, an accelerated mode trigger being set for an accelerated guest determination SIE instruction execution means that parallel use of hardware facilities is permitted.
21. In a data processing system as defined in Claim 20, the CP further comprising:

limit registering means being enabled by the set state of the accelerated mode trigger for receiving an upper limit main storage address as the upper boundary allocated in real storage to an accelerated guest, means for comparing each translated address requested by the accelerated guest with the upper limit main storage address in the limit registering means, means for generating an addressing exception signal when the translated address is greater than the upper limit main storage address, whereby limit checking is executed for accelerated guest translated addresses with no CP performance degradation caused by either the limit checking or accelerated guest address translation.
22. In a data processing system as defined in Claim 21, the CP further comprising:

prefix registering means in the CP for receiving a prefix value, means for transferring into the prefix registering means an accelerated guest prefix value from main storage in response to the set state of the accelerated mode trigger in order to prefix storage requests by the accelerated guest, whereby any accelerated guest prefixing operation has no performance degradation in comparison to a host prefixing operation in the CP, even though the CP is a uniprocessor.
23. In a data processing system as defined in Claim 22, the CP further comprising:

means for enabling the accelerated mode trigger to be set to the accelerated guest state only if the SIE instruction execution means accesses a preferred guest indicating field.
24. In a data processing system as defined in Claim 11, 12, or 14, the system further comprising:

a plurality of CPs in which each CP is identical to the CP to provide a multiprocessing system, means connecting the plurality of CPs to the same main storage, whereby any guest program may be executed on any CP in the multiprocessing system.
25. In a data processing system as defined in Claim 22 or 23, the system further comprising:

a plurality of CPs in which each CP is identical to the CP to provide a multiprocessing system, means connecting the plurality of CPs to the same main storage, whereby any guest program may be executed on any CP in the multiprocessing system.
CA000400593A 1981-06-15 1982-04-07 Guest architectural support in a computer system Expired CA1176377A (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US06/273,532 US4456954A (en) 1981-06-15 1981-06-15 Virtual machine system with guest architecture emulation using hardware TLB's for plural level address translations
US273,532 1981-06-15

Publications (1)

Publication Number Publication Date
CA1176377A true CA1176377A (en) 1984-10-16

Family

ID=23044327

Family Applications (1)

Application Number Title Priority Date Filing Date
CA000400593A Expired CA1176377A (en) 1981-06-15 1982-04-07 Guest architectural support in a computer system

Country Status (7)

Country Link
US (1) US4456954A (en)
EP (1) EP0067344B1 (en)
JP (1) JPS57212680A (en)
AU (1) AU547403B2 (en)
CA (1) CA1176377A (en)
DE (1) DE3275771D1 (en)
ES (1) ES8304331A1 (en)

Families Citing this family (160)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS58180990U (en) * 1982-05-28 1983-12-02 古河鉱業株式会社 Rod for mudstone drilling
US5109522A (en) * 1983-06-02 1992-04-28 Amdahl Corporation Data processing system with logical processing facility supporting a plurality of system control programs for production operation
JPS608971A (en) * 1983-06-29 1985-01-17 Toshiba Corp Central processing unit
US4612612A (en) * 1983-08-30 1986-09-16 Amdahl Corporation Virtually addressed cache
JPS6057438A (en) * 1983-09-08 1985-04-03 Hitachi Ltd Virtual computer system controller
US4564903A (en) * 1983-10-05 1986-01-14 International Business Machines Corporation Partitioned multiprocessor programming system
JPS60107156A (en) * 1983-11-16 1985-06-12 Hitachi Ltd Data processing system
US4779188A (en) * 1983-12-14 1988-10-18 International Business Machines Corporation Selective guest system purge control
CA1213986A (en) * 1983-12-14 1986-11-12 Thomas O. Curlee, Iii Selective guest system purge control
JPS60142451A (en) * 1983-12-29 1985-07-27 Fujitsu Ltd Address conversion control system
US5392409A (en) * 1984-01-18 1995-02-21 Hitachi, Ltd. I/O execution method for a virtual machine system and system therefor
JPH0619747B2 (en) * 1984-01-18 1994-03-16 株式会社日立製作所 I / O instruction execution method, I / O interrupt processing method, and computer system using them
GB8405491D0 (en) * 1984-03-02 1984-04-04 Hemdal G Computers
EP0159699A3 (en) * 1984-04-23 1988-09-28 Nec Corporation A data processor executing microprograms according to a plurality of system architectures
US4603235A (en) * 1984-05-21 1986-07-29 Gte Communication Systems Corporation Dynamic event selection network
US4792895A (en) * 1984-07-30 1988-12-20 International Business Machines Corp. Instruction processing in higher level virtual machines by a real machine
US4695950A (en) * 1984-09-17 1987-09-22 International Business Machines Corporation Fast two-level dynamic address translation method and means
JPH0652511B2 (en) * 1984-12-14 1994-07-06 株式会社日立製作所 Address conversion method for information processing equipment
US4975836A (en) * 1984-12-19 1990-12-04 Hitachi, Ltd. Virtual computer system
US4674038A (en) * 1984-12-28 1987-06-16 International Business Machines Corporation Recovery of guest virtual machines after failure of a host real machine
JPS625441A (en) * 1985-02-18 1987-01-12 Nec Corp Information processor
JPS61190638A (en) * 1985-02-20 1986-08-25 Hitachi Ltd File control system for virtual computer
US5255384A (en) * 1985-02-22 1993-10-19 Intergraph Corporation Memory address translation system having modifiable and non-modifiable translation mechanisms
US4649479A (en) * 1985-02-28 1987-03-10 International Business Machines Corp. Device driver and adapter binding technique
JPS61206043A (en) * 1985-03-11 1986-09-12 Hitachi Ltd Interruption control method in virtual computer system
JPS61206057A (en) * 1985-03-11 1986-09-12 Hitachi Ltd Address converting device
JP2539357B2 (en) * 1985-03-15 1996-10-02 株式会社日立製作所 Data processing device
US4737909A (en) * 1985-04-01 1988-04-12 National Semiconductor Corp. Cache memory address apparatus
US4660144A (en) * 1985-05-23 1987-04-21 International Business Machines Corp. Adjunct machine
JPH0685156B2 (en) * 1985-05-24 1994-10-26 株式会社日立製作所 Address translator
JPH0792761B2 (en) * 1985-07-31 1995-10-09 株式会社日立製作所 Input / output control method for virtual computer system
JPH0658649B2 (en) * 1985-10-28 1994-08-03 株式会社日立製作所 Area management method in virtual memory device
JPS62154037A (en) * 1985-12-26 1987-07-09 Hitachi Ltd Virtual computer supervising and controlling system
US4742447A (en) * 1986-01-16 1988-05-03 International Business Machines Corporation Method to control I/O accesses in a multi-tasking virtual memory virtual machine type data processing system
JPH0658650B2 (en) * 1986-03-14 1994-08-03 株式会社日立製作所 Virtual computer system
US5349672A (en) * 1986-03-17 1994-09-20 Hitachi, Ltd. Data processor having logical address memories and purge capabilities
US4862347A (en) * 1986-04-22 1989-08-29 International Business Machine Corporation System for simulating memory arrays in a logic simulation machine
US5113517A (en) * 1986-04-28 1992-05-12 Xerox Corporation Concurrent display of data from two different processors each having different display font and user interface for controlling transfer of converted font data therebetween
US4920481A (en) * 1986-04-28 1990-04-24 Xerox Corporation Emulation with display update trapping
US5153577A (en) * 1986-04-28 1992-10-06 Xerox Corporation Mapping character color attributes into grey pixel patterns
US4899136A (en) * 1986-04-28 1990-02-06 Xerox Corporation Data processor having a user interface display with metaphoric objects
US4937036A (en) * 1986-04-28 1990-06-26 Xerox Corporation Concurrent display of data from two different display processors and user interface therefore
US4939507A (en) * 1986-04-28 1990-07-03 Xerox Corporation Virtual and emulated objects for use in the user interface of a display screen of a display processor
US5088033A (en) * 1986-04-28 1992-02-11 Xerox Corporation Data processing system emulation in a window with a coprocessor and I/O emulation
US4797814A (en) * 1986-05-01 1989-01-10 International Business Machines Corporation Variable address mode cache
US5038281A (en) * 1986-09-19 1991-08-06 International Business Machines Corporation Acceleration of system interrupts between operating systems in guest-host relationship
JPH0810437B2 (en) * 1987-05-11 1996-01-31 株式会社日立製作所 Guest execution control method for virtual machine system
US5218712A (en) * 1987-07-01 1993-06-08 Digital Equipment Corporation Providing a data processor with a user-mode accessible mode of operations in which the processor performs processing operations without interruption
JP2523653B2 (en) * 1987-07-08 1996-08-14 株式会社日立製作所 Virtual computer system
JP2510605B2 (en) * 1987-07-24 1996-06-26 株式会社日立製作所 Virtual computer system
US4843541A (en) * 1987-07-29 1989-06-27 International Business Machines Corporation Logical resource partitioning of a data processing system
US4975869A (en) * 1987-08-06 1990-12-04 International Business Machines Corporation Fast emulator using slow processor
JP2615103B2 (en) * 1987-12-11 1997-05-28 株式会社日立製作所 Virtual computer system
US4943913A (en) * 1988-02-10 1990-07-24 International Business Machines Corporation Operating system accessing control blocks by using home address space segment table to control instruction and operand fetch and store operations
JPH01255945A (en) * 1988-04-06 1989-10-12 Hitachi Ltd Address converter in virtual computer
JPH02202652A (en) * 1989-02-01 1990-08-10 Hitachi Ltd Multiple virtical storage control system
IT1228728B (en) * 1989-03-15 1991-07-03 Bull Hn Information Syst MULTIPROCESSOR SYSTEM WITH GLOBAL DATA REPLICATION AND TWO LEVELS OF ADDRESS TRANSLATION UNIT.
US5293612A (en) * 1989-05-11 1994-03-08 Tandem Computers Incorporated Selective dump method and apparatus
US5155809A (en) * 1989-05-17 1992-10-13 International Business Machines Corp. Uncoupling a central processing unit from its associated hardware for interaction with data handling apparatus alien to the operating system controlling said unit and hardware
JP2825550B2 (en) * 1989-09-21 1998-11-18 株式会社日立製作所 Multiple virtual space address control method and computer system
US5469556A (en) * 1989-12-12 1995-11-21 Harris Corporation Resource access security system for controlling access to resources of a data processing system
JPH03217949A (en) * 1990-01-23 1991-09-25 Hitachi Ltd Computer system
JP2708608B2 (en) * 1990-05-25 1998-02-04 富士通株式会社 Virtual machine IPL processing method
JP2839201B2 (en) * 1990-07-30 1998-12-16 株式会社日立製作所 Virtual computer system
US5283876A (en) * 1990-10-05 1994-02-01 Bull Hn Information Systems Inc. Virtual memory unit utilizing set associative memory structure and state machine control sequencing with selective retry
US5317754A (en) * 1990-10-23 1994-05-31 International Business Machines Corporation Method and apparatus for enabling an interpretive execution subset
US5381535A (en) * 1990-10-24 1995-01-10 International Business Machines Corporation Data processing control of second-level quest virtual machines without host intervention
US5317705A (en) * 1990-10-24 1994-05-31 International Business Machines Corporation Apparatus and method for TLB purge reduction in a multi-level machine system
US5412787A (en) * 1990-11-21 1995-05-02 Hewlett-Packard Company Two-level TLB having the second level TLB implemented in cache tag RAMs
EP0532744B1 (en) * 1991-03-07 1996-12-18 Digital Equipment Corporation Improved software debugging system and method especially adapted for code debugging within a multi-architecture environment
US5652869A (en) * 1991-03-07 1997-07-29 Digital Equipment Corporation System for executing and debugging multiple codes in a multi-architecture environment using jacketing means for jacketing the cross-domain calls
US5341485A (en) * 1991-05-07 1994-08-23 International Business Machines Corporation Multiple virtual address translation per computer cycle
JPH0512126A (en) * 1991-07-05 1993-01-22 Hitachi Ltd Device and method for address conversion for virtual computer
US5428757A (en) * 1992-04-29 1995-06-27 International Business Machines Corporation Method for reducing translation look aside buffer purges in a multitasking system
JPH0695898A (en) * 1992-09-16 1994-04-08 Hitachi Ltd Control method for virtual computer and virtual computer system
JPH06348584A (en) * 1993-06-01 1994-12-22 Internatl Business Mach Corp <Ibm> Data processing system
US5584042A (en) * 1993-06-01 1996-12-10 International Business Machines Corporation Dynamic I/O data address relocation facility
US6684261B1 (en) * 1993-07-19 2004-01-27 Object Technology Licensing Corporation Object-oriented operating system
US5404529A (en) * 1993-07-19 1995-04-04 Taligent, Inc. Object-oriented interprocess communication system interface for a procedural operating system
US5379432A (en) 1993-07-19 1995-01-03 Taligent, Inc. Object-oriented interface for a procedural operating system
US5455951A (en) * 1993-07-19 1995-10-03 Taligent, Inc. Method and apparatus for running an object-oriented program on a host computer with a procedural operating system
US5555385A (en) * 1993-10-27 1996-09-10 International Business Machines Corporation Allocation of address spaces within virtual machine compute system
US5781750A (en) * 1994-01-11 1998-07-14 Exponential Technology, Inc. Dual-instruction-set architecture CPU with hidden software emulation mode
JP2806778B2 (en) * 1994-01-28 1998-09-30 甲府日本電気株式会社 Conversion index buffer clear instruction processing method
US5664156A (en) * 1994-09-16 1997-09-02 Philips Electronics North America Corporation Microcontroller with a reconfigurable program status word
US5555414A (en) * 1994-12-14 1996-09-10 International Business Machines Corporation Multiprocessing system including gating of host I/O and external enablement to guest enablement at polling intervals
JPH08286932A (en) * 1995-04-11 1996-11-01 Hitachi Ltd Parallel execution control method for job
JP3451595B2 (en) * 1995-06-07 2003-09-29 インターナショナル・ビジネス・マシーンズ・コーポレーション Microprocessor with architectural mode control capable of supporting extension to two distinct instruction set architectures
US6356918B1 (en) 1995-07-26 2002-03-12 International Business Machines Corporation Method and system for managing registers in a data processing system supports out-of-order and speculative instruction execution
US5764944A (en) * 1995-09-08 1998-06-09 United Microelectronics Corp. Method and apparatus for TLB invalidation mechanism for protective page fault
JPH0981459A (en) * 1995-09-19 1997-03-28 Hitachi Ltd Address conversion buffer device
US5787492A (en) * 1996-04-09 1998-07-28 International Business Machines Corporation Address limit check apparatus with conditional carry logic
US6047388A (en) * 1997-04-09 2000-04-04 International Business Machines Corporation Method and apparatus for processing an invalid address request
US6668314B1 (en) * 1997-06-24 2003-12-23 Hewlett-Packard Development Company, L.P. Virtual memory translation control by TLB purge monitoring
US6397242B1 (en) 1998-05-15 2002-05-28 Vmware, Inc. Virtualization system including a virtual machine monitor for a computer with a segmented architecture
US7516453B1 (en) * 1998-10-26 2009-04-07 Vmware, Inc. Binary translator with precise exception synchronization mechanism
JP2001051900A (en) * 1999-08-17 2001-02-23 Hitachi Ltd Information processing unit and processor for virtual computer system
FR2797970A1 (en) * 1999-08-31 2001-03-02 Koninkl Philips Electronics Nv ADDRESSING A MEMORY
US6415379B1 (en) * 1999-10-13 2002-07-02 Transmeta Corporation Method and apparatus for maintaining context while executing translated instructions
US6651132B1 (en) * 2000-07-17 2003-11-18 Microsoft Corporation System and method for emulating the operation of a translation look-aside buffer
US7149878B1 (en) 2000-10-30 2006-12-12 Mips Technologies, Inc. Changing instruction set architecture mode by comparison of current instruction execution address with boundary address register values
US6742180B1 (en) * 2000-10-30 2004-05-25 Microsoft Corporation System and method providing seamless transition of operating system environment
US6907600B2 (en) 2000-12-27 2005-06-14 Intel Corporation Virtual translation lookaside buffer
DE10104984A1 (en) * 2001-02-03 2002-08-08 Hejoe Ges Fuer Netzwerk Und Sy System and control method for networked computers is such that the computers have a base operating system that is common with the server system, with a virtual computer running on top of the base system to suit user requirements
US7107439B2 (en) * 2001-08-10 2006-09-12 Mips Technologies, Inc. System and method of controlling software decompression through exceptions
US7069413B1 (en) 2003-01-29 2006-06-27 Vmware, Inc. Method and system for performing virtual to physical address translations in a virtual machine monitor
US6981125B2 (en) * 2003-04-22 2005-12-27 International Business Machines Corporation Method and apparatus for managing shared virtual storage in an information handling system
WO2004104825A1 (en) 2003-05-15 2004-12-02 Applianz Technologies, Inc. Systems and methods of creating and accessing software simulated computers
US20050034125A1 (en) * 2003-08-05 2005-02-10 Logicube, Inc. Multiple virtual devices
US20060005190A1 (en) * 2004-06-30 2006-01-05 Microsoft Corporation Systems and methods for implementing an operating system in a virtual machine environment
US9058292B2 (en) 2004-12-29 2015-06-16 Intel Corporation System and method for one step address translation of graphics addresses in virtualization
US7886126B2 (en) * 2005-01-14 2011-02-08 Intel Corporation Extended paging tables to map guest physical memory addresses from virtual memory page tables to host physical memory addresses in a virtual machine system
US7496495B2 (en) * 2005-05-12 2009-02-24 Microsoft Corporation Virtual operating system device communication relying on memory access violations
US20070006178A1 (en) * 2005-05-12 2007-01-04 Microsoft Corporation Function-level just-in-time translation engine with multiple pass optimization
US7299337B2 (en) * 2005-05-12 2007-11-20 Traut Eric P Enhanced shadow page table algorithms
US7868897B2 (en) 2006-06-30 2011-01-11 Intel Corporation Apparatus and method for memory address re-mapping of graphics data
US7881921B1 (en) * 2006-07-05 2011-02-01 Synopsys, Inc. Caching information to map simulation addresses to host addresses in computer system simulations
US7555628B2 (en) * 2006-08-15 2009-06-30 Intel Corporation Synchronizing a translation lookaside buffer to an extended paging table
US8819647B2 (en) * 2008-01-25 2014-08-26 International Business Machines Corporation Performance improvements for nested virtual machines
US7530106B1 (en) 2008-07-02 2009-05-05 Kaspersky Lab, Zao System and method for security rating of computer processes
US8275971B2 (en) 2008-08-27 2012-09-25 International Business Machines Corporation Method and apparatus for managing software controlled cache of translating the physical memory access of a virtual machine between different levels of translation entities
US7904914B2 (en) * 2008-09-30 2011-03-08 Microsoft Corporation On-the-fly replacement of physical hardware with emulation
US10802990B2 (en) * 2008-10-06 2020-10-13 International Business Machines Corporation Hardware based mandatory access control
US8301863B2 (en) * 2008-11-17 2012-10-30 International Business Machines Corporation Recursive logical partition real memory map
US8135937B2 (en) * 2008-11-17 2012-03-13 International Business Machines Corporation Logical partition memory
US10255463B2 (en) 2008-11-17 2019-04-09 International Business Machines Corporation Secure computer architecture
US8387031B2 (en) * 2009-01-23 2013-02-26 International Business Machines Corporation Providing code improvements for nested virtual machines
US8286164B2 (en) * 2009-08-07 2012-10-09 International Business Machines Corporation Secure recursive virtualization
US20130115844A1 (en) 2010-07-16 2013-05-09 Nitto Denko Corporation Electrically insulating resin composition and laminate sheet
US9251091B2 (en) 2012-06-15 2016-02-02 International Business Machines Corporation Translation look-aside table management
US9355032B2 (en) 2012-10-08 2016-05-31 International Business Machines Corporation Supporting multiple types of guests by a hypervisor
US9600419B2 (en) 2012-10-08 2017-03-21 International Business Machines Corporation Selectable address translation mechanisms
US9355040B2 (en) 2012-10-08 2016-05-31 International Business Machines Corporation Adjunct component to provide full virtualization using paravirtualized hypervisors
US9280488B2 (en) 2012-10-08 2016-03-08 International Business Machines Corporation Asymmetric co-existent address translation structure formats
US9740624B2 (en) 2012-10-08 2017-08-22 International Business Machines Corporation Selectable address translation mechanisms within a partition
US9348757B2 (en) 2012-10-08 2016-05-24 International Business Machines Corporation System supporting multiple partitions with differing translation formats
US9223574B2 (en) * 2014-03-27 2015-12-29 International Business Machines Corporation Start virtual execution instruction for dispatching multiple threads in a computer
US9213569B2 (en) 2014-03-27 2015-12-15 International Business Machines Corporation Exiting multiple threads in a computer
US9772867B2 (en) 2014-03-27 2017-09-26 International Business Machines Corporation Control area for managing multiple threads in a computer
US9195493B2 (en) * 2014-03-27 2015-11-24 International Business Machines Corporation Dispatching multiple threads in a computer
US10353680B2 (en) 2014-07-25 2019-07-16 Intel Corporation System converter that implements a run ahead run time guest instruction conversion/decoding process and a prefetching process where guest code is pre-fetched from the target of guest branches in an instruction sequence
US11281481B2 (en) 2014-07-25 2022-03-22 Intel Corporation Using a plurality of conversion tables to implement an instruction set agnostic runtime architecture
US9733909B2 (en) 2014-07-25 2017-08-15 Intel Corporation System converter that implements a reordering process through JIT (just in time) optimization that ensures loads do not dispatch ahead of other loads that are to the same address
US20160026487A1 (en) 2014-07-25 2016-01-28 Soft Machines, Inc. Using a conversion look aside buffer to implement an instruction set agnostic runtime architecture
US9389897B1 (en) 2014-12-18 2016-07-12 International Business Machines Corporation Exiting multiple threads of a simulation environment in a computer
JP6678185B2 (en) * 2015-04-10 2020-04-08 グーグル エルエルシー Binary conversion to native client
US10169243B2 (en) 2016-07-18 2019-01-01 International Business Machines Corporation Reducing over-purging of structures associated with address translation
US10176111B2 (en) 2016-07-18 2019-01-08 International Business Machines Corporation Host page management using active guest page table indicators
US10241924B2 (en) 2016-07-18 2019-03-26 International Business Machines Corporation Reducing over-purging of structures associated with address translation using an array of tags
US10282305B2 (en) 2016-07-18 2019-05-07 International Business Machines Corporation Selective purging of entries of structures associated with address translation in a virtualized environment
US10180909B2 (en) 2016-07-18 2019-01-15 International Business Machines Corporation Host-based resetting of active use of guest page table indicators
US10176006B2 (en) 2016-07-18 2019-01-08 International Business Machines Corporation Delaying purging of structures associated with address translation
US10223281B2 (en) 2016-07-18 2019-03-05 International Business Machines Corporation Increasing the scope of local purges of structures associated with address translation
US10248573B2 (en) 2016-07-18 2019-04-02 International Business Machines Corporation Managing memory used to back address translation structures
US10168902B2 (en) 2016-07-18 2019-01-01 International Business Machines Corporation Reducing purging of structures associated with address translation
US10162764B2 (en) 2016-07-18 2018-12-25 International Business Machines Corporation Marking page table/page status table entries to indicate memory used to back address translation structures
US10802986B2 (en) 2016-07-18 2020-10-13 International Business Machines Corporation Marking to indicate memory used to back address translation structures
US10176110B2 (en) 2016-07-18 2019-01-08 International Business Machines Corporation Marking storage keys to indicate memory used to back address translation structures
US11347869B2 (en) * 2019-03-08 2022-05-31 International Business Machines Corporation Secure interface control high-level page management
US11403409B2 (en) 2019-03-08 2022-08-02 International Business Machines Corporation Program interruptions for page importing/exporting

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR130806A (en) * 1973-11-21
US4351024A (en) * 1975-04-21 1982-09-21 Honeywell Information Systems Inc. Switch system base mechanism
FR2400729A1 (en) * 1977-08-17 1979-03-16 Cii Honeywell Bull DEVICE FOR THE TRANSFORMATION OF VIRTUAL ADDRESSES INTO PHYSICAL ADDRESSES IN A DATA PROCESSING SYSTEM
US4285040A (en) * 1977-11-04 1981-08-18 Sperry Corporation Dual mode virtual-to-real address translation mechanism
US4241401A (en) * 1977-12-19 1980-12-23 Sperry Corporation Virtual address translator utilizing interrupt level code
JPS5576447A (en) * 1978-12-01 1980-06-09 Fujitsu Ltd Address control system for software simulation
US4253145A (en) * 1978-12-26 1981-02-24 Honeywell Information Systems Inc. Hardware virtualizer for supporting recursive virtual computer systems on a host computer system
JPS55112651A (en) * 1979-02-21 1980-08-30 Fujitsu Ltd Virtual computer system

Also Published As

Publication number Publication date
DE3275771D1 (en) 1987-04-23
JPS57212680A (en) 1982-12-27
EP0067344A3 (en) 1984-05-02
US4456954A (en) 1984-06-26
ES511808A0 (en) 1983-02-16
JPS6122825B2 (en) 1986-06-03
EP0067344B1 (en) 1987-03-18
AU8341582A (en) 1982-12-23
EP0067344A2 (en) 1982-12-22
ES8304331A1 (en) 1983-02-16
AU547403B2 (en) 1985-10-17

Similar Documents

Publication Publication Date Title
CA1176377A (en) Guest architectural support in a computer system
US4695950A (en) Fast two-level dynamic address translation method and means
US5574936A (en) Access control mechanism controlling access to and logical purging of access register translation lookaside buffer (ALB) in a computer system
EP1701269B1 (en) Invalidating storage, clearing buffer entries
US4779188A (en) Selective guest system purge control
US5555395A (en) System for memory table cache reloads in a reduced number of cycles using a memory controller to set status bits in the main memory table
US5339417A (en) Computer system with two levels of guests
US11474956B2 (en) Memory protection unit using memory protection table stored in memory system
US5615327A (en) Apparatus and method for controlling the initiating of a synchronization protocol for synchronizing a plurality of processors in a multi-processor system
EP0145960B1 (en) Selective guest system purge control
US5280592A (en) Domain interlock
EP0619899B1 (en) Software control of hardware interruptions
EP3830719B1 (en) Binary search procedure for control table stored in memory system
JPH06332803A (en) Tlb control method in virtual computer system
EP0550285A2 (en) Machine with two units of operation
JPS63286944A (en) Nullification system for address translation buffer

Legal Events

Date Code Title Description
MKEC Expiry (correction)
MKEX Expiry