Update cwsdpmi and the DOS README

10 files changed, 309 insertions, 834 deletions

diff --git a/cwsdpmi/bin/cwsdpmi.doc b/cwsdpmi/bin/cwsdpmi.doc
index de49034..13a7b97 100644
--- a/cwsdpmi/bin/cwsdpmi.doc
+++ b/cwsdpmi/bin/cwsdpmi.doc
@@ -1,173 +1,196 @@
-CWSDPMI is Copyright (C) 1995-2000  Charles W Sandmann (sandmann@clio.rice.edu)

-                                    1206 Braelinn, Sugar Land, TX 77479

-

-This is release 5.  The files in this binary distribution may be redistributed

-under the GPL (with source) or without the source code provided:

-

-* CWSDPMI.EXE or CWSDPR0.EXE are not modified in any way except via CWSPARAM.

-

-* CWSDSTUB.EXE internal contents are not modified in any way except via

-  CWSPARAM or STUBEDIT.  It may have a COFF image plus data appended to it.

-

-* Notice to users that they have the right to receive the source code and/or

-  binary updates for CWSDPMI.  Distributors should indicate a site for the

-  source in their documentation.

-

--------------------------------------------------------------------------------

-

-CWSDPMI was written to provide DPMI services for V2 of DJGPP.  It currently

-does not support 16-bit DPMI applications, or DPMI applications requiring a

-built in extender.  It does support virtual memory and hardware interrupt

-reflection from real mode to protected mode.  DJGPP V1.1x and RSX applications

-will also run using this server, which can be used to provide enhanced control

-over hardware interrupts.  Some DPMI 1.0 extensions (0x506, 0x507, 0x508) have

-been implemented.

-

-CWSDPR0.EXE is an alternate version which runs at ring 0 with virtual memory

-disabled.  It may be used if access to ring-0 features are desired.  It

-currently does not switch stacks on HW interrupts, so some DJGPP features

-such as SIGINT and SIGFPE are not supported and will generate a double fault

-or stack fault error (to be fixed someday).

-

-CWSDSTUB.EXE is a stub loader image for DJGPP which includes CWSDPMI.  This

-allows single executable image distributions.  You can use the EXE2COFF

-program and COPY /B CWSDSTUB.EXE+yourimage yourimage.exe to create a

-standalone executable image.

-

-Some of the internal tuning and configuration parameters may be modified

-in the image using CWSPARAM.EXE (see CWSPARAM.DOC).

-

-If you want to use CWSDPMI with DJGPP, you expand the distribution into the

-DJGPP directory tree.  CWSDPMI.EXE will be put in the BIN directory with your

-DJGPP images and it will automatically be loaded when they run.

-

-Directions for use (server can be used in either of two different ways):

-

-1) "cwsdpmi" alone with no parameters will terminate and stay resident

-   FOR A SINGLE DPMI PROCESS.  This means it unloads itself when your

-   DPMI application exits.  This mode is useful in software which needs

-   DPMI services, since CWSDPMI can be exec'ed and then will unload on exit.

-

-2) "cwsdpmi -p" will terminate and stay resident until you remove it.

-   It can be loaded into UMBs with LH.  "cwsdpmi -u" will unload the TSR.

-

-3) The file used for virtual memory swapping, if desired, is controlled

-   by the "-sc:\cwsdpmi.swp" syntax on the command line.  You must specify

-   either a file with full disk/directory syntax, or "-s-" which disables

-   virtual memory.

-

-4) The default swap file name is c:\cwsdpmi.swp, but this can be changed

-   with the CWSPARAM image, as can some other parameters.

-

-5) You can disable the DPMI 1.0 extensions by starting the image with the

-   "cwsdpmi -x" syntax.  This feature allows you to run programs developed

-   under other DPMI providers which do not behave properly with these

-   extensions enabled (typically use of NULL pointers).

-

-I would like to give special thanks to DJ Delorie who wrote the original

-GO32 code on which CWSDPMI is based.  Morten Welinder also provided and

-improved much of the code in this program.

-

--------------------------------------------------------------------------------

-

-This section contains a list of the error messages you might see out of

-CWSDPMI and some details on what they mean.

-

-Exceptions are only handled by CWSDPMI if the application does not establish

-an exception handler, exceptions nest 5 deep, or the error is particularly bad:

-

-"Page fault" -

-  1) an illegal page fault happens in a RMCB or HW interrupt, (lock all pages!)

-  2) all available pages have been locked,

-  3) the application is using non-committed pages for null pointer protection.

-"Double Fault" - multiple exceptions occurred

-"Invalid TSS" - typically due to RMCB or HW interrupt being called after the

-   selectors/memory have been deallocated (remember to reset the mouse)

-"General Protection Fault" - bad parameter sent to a DPMI call

-

-"80386 required."

-

-Since 80286 and lesser processors don't have the hardware necessary to

-run CWSDPMI.  No workaround, upgrade.

-

-"DOS 3 required."

-

-A few interrupts are used which need DOS 3.0 or higher.  I don't expect to

-ever see this message, since 80386 machines were introduced after DOS 3.0

-and that check is made first.

-

-"CWSDPMI V0.90+ (r5) Copyright (C) 2000 CW Sandmann  ABSOLUTELY NO WARRANTY"

-

-An informational message displayed if the program is not run in one-pass mode.

-

-"Protected mode not accessible."

-

-This message should only be displayed if running CWSDPMI in a protected

-environment with no access to protected mode.  In this case, DPMI should

-already be available and CWSDPMI would not be needed.  This might happen if

-a 16-bit DPMI client is loaded and a DJGPP image attempts to load CWSDPMI

-to provide 32-bit DPMI services under Windows.

-

-"Warning: cannot open swap file c:\cwsdpmi.swp"

-

-Maybe you are out of file handles, or the swap file name is incorrectly

-specified in the image (change the name with cwsparam).

-

-"No swap space!"

-

-This message means you tried to use more paging file than CWSDPMI was

-configured to handle.  Since this is protected against in the memory

-allocation code, you should never see this message.

-

-"Swap disk full!"

-

-This means the paging file could not be expanded when trying to page

-memory out to disk.  This would normally not be seen, unless you are

-writing output to the same disk which holds the paging file.  Decrease

-the amount of memory your DPMI application is using or free up disk space.

-

-"Interrupt 0x??"

-

-Your application tried to call an interrupt from protected mode which

-normally shouldn't be called (something like a data pointer).  If the

-request was allowed to continue it would likely hang your machine.  If you

-see this message and think the interrupt should be allowed to continue, let

-me know.

-

-"Error: Using XMS switched CPU into V86 mode."

-

-This message might be seen if you have your memory manager in AUTO mode.  The

-only workaround in this case is to stop using AUTO mode.

-

-"Error: could not allocate page table memory"

-

-The page table memory (a minimum of 16Kb) is allocated from conventional

-memory (either in the 640Kb region or UMBs).  If CWSDPMI cannot allocate the

-minimum necessary memory, you would see this message.  Free up some

-conventional memory.  You may also see this message if a page directory needs

-to be faulted in, and there are no available pages.  This means too many pages

-have been locked for the allocated page tables available.  While CWSDPMI

-tries to dynamically allocate these if needed, this effort failed.  You need

-to increase the number of page tables with CWSPARAM, or increase the amount

-of free conventional memory if it is low.  If the application which calls

-CWSDPMI internally manages all the DOS memory, the page tables may need to

-be pre-allocated at DPMI startup time (if this is needed, try using the

-run option flag 2 in cwsparam).

-

-"16-bit DPMI unsupported."

-

-CWSDPMI is a 32-bit only DPMI server.  Ideally, on the request to enter DPMI's

-PM with a 16-bit request, we would just fail the call setting the carry bit

-like the DPMI specification describes.  Some buggy 16-bit compiler tools don't

-check the return status and will hang the machine in this case.  So, I issue

-an error message and exit the image instead.

-

-"Descriptors exhausted."

-

-An attempt to nest a DPMI client failed in the setup phase due to insufficient

-free selectors in the LDT.

-

-"CWSDPMI not removed"

-

-When the -u parameter is specified, if DPMI is not detected this message is

-printed.  Informational.

+CWSDPMI is Copyright (C) 1995-2010  Charles W Sandmann (cwsdpmi@earthlink.net)
+                                    1206 Braelinn, Sugar Land, TX 77479
+
+This is release 7.  The files in this binary distribution may be redistributed
+under the GPL (with source) or without the source code provided:
+
+* CWSDPMI.EXE or CWSDPR0.EXE are not modified in any way except via CWSPARAM.
+
+* CWSDSTUB.EXE internal contents are not modified in any way except via
+  CWSPARAM or STUBEDIT.  It may have a COFF image plus data appended to it.
+
+* Notice to users that they have the right to receive the source code and/or
+  binary updates for CWSDPMI.  Distributors should indicate a site for the
+  source in their documentation.
+
+-------------------------------------------------------------------------------
+
+The purpose of this release is to allow high performance access to large
+amounts of physical memory (up to 4GB).  Memory above 32-bit address space
+(common on machines over 3GB physical memory) will be mapped to fill the
+address hole to provide a full 4GB memory.  4MB pages will be automatically
+used for large memory requests.  Release 5 works well for systems using 256MB
+or less of memory; release 7 is useful for quick access to over 500MB.
+
+Known limitations with 4MB pages:
+ 1) 4Mb regions are treated as locked (non-pageable).  Virtual memory may be
+    less effective.
+ 2) DPMI 1.0 extensions to manage page attributes will not work on 4MB pages.
+ 3) DPMI function 0x503 to resize memory is not supported if it requires the
+    memory zone to be moved in physical memory and it contains a 4Mb page.
+Note: CWSPARAM bit 3, (bitmask 8) can be used to suppress using 4Mb pages.
+
+CWSDPMI was written to provide DPMI services for V2 of DJGPP.  It
+does not support 16-bit DPMI applications, or DPMI applications requiring a
+built in extender.  It does support virtual memory and hardware interrupt
+reflection from real mode to protected mode.  DJGPP V1.1x and RSX applications
+will also run using this server, which can be used to provide enhanced control
+over hardware interrupts.  Some DPMI 1.0 extensions (0x506, 0x507, 0x508) have
+been implemented.
+
+CWSDPR0.EXE is an alternate version which runs at ring 0 with virtual memory
+disabled.  It may be used if access to ring-0 features are desired.  It
+currently does not switch stacks on HW interrupts, so some DJGPP features
+such as SIGINT and SIGFPE are not supported and will generate a double fault
+or stack fault error (to be fixed someday).
+
+CWSDSTUB.EXE is a stub loader image for DJGPP which includes CWSDPMI.  This
+allows single executable image distributions.  You can use the EXE2COFF
+program and COPY /B CWSDSTUB.EXE+yourimage yourimage.exe to create a
+standalone executable image.  CWSDSTR0.EXE is the ring 0 stub.
+
+Some of the internal tuning and configuration parameters may be modified
+in the image using CWSPARAM.EXE (see CWSPARAM.DOC).
+
+If you want to use CWSDPMI with DJGPP, you expand the distribution into the
+DJGPP directory tree.  CWSDPMI.EXE will be put in the BIN directory with your
+DJGPP images and it will automatically be loaded when they run.
+
+Directions for use (server can be used in either of two different ways):
+
+1) "cwsdpmi" alone with no parameters will terminate and stay resident
+   FOR A SINGLE DPMI PROCESS.  This means it unloads itself when your
+   DPMI application exits.  This mode is useful in software which needs
+   DPMI services, since CWSDPMI can be exec'ed and then will unload on exit.
+
+2) "cwsdpmi -p" will terminate and stay resident until you remove it.
+   It can be loaded into UMBs with LH.  "cwsdpmi -u" will unload the TSR.
+
+3) The file used for virtual memory swapping, if desired, is controlled
+   by the "-sc:\cwsdpmi.swp" syntax on the command line.  You must specify
+   either a file with full disk/directory syntax, or "-s-" which disables
+   virtual memory.  If -s is used, it must be the last option on the command
+   line, since all characters after -s will be treated as the swap file name.
+
+4) The default swap file name is c:\cwsdpmi.swp, but this can be changed
+   with the CWSPARAM image, as can some other parameters.
+
+5) You can disable the DPMI 1.0 extensions by starting the image with the
+   "cwsdpmi -x" syntax.  This feature allows you to run programs developed
+   under other DPMI providers which do not behave properly with these
+   extensions enabled (typically use of NULL pointers).
+
+If you are using an EMM provider for UMBs, CWSDPMI will need to use VCPI.
+VCPI memory prevents usage of 4MB pages, so for best performance with an EMM
+provider I recommend suppressing VCPI memory.  For example, with JEMM386
+this can be done with:  JEMM386 LOAD MAX=0
+
+I would like to give special thanks to DJ Delorie who wrote the original
+GO32 code on which CWSDPMI is based.  Morten Welinder also provided and
+improved much of the code in this program.
+
+-------------------------------------------------------------------------------
+
+This section contains a list of the error messages you might see out of
+CWSDPMI and some details on what they mean.
+
+Exceptions are only handled by CWSDPMI if the application does not establish
+an exception handler, exceptions nest 5 deep, or the error is particularly bad:
+
+"Page fault" -
+  1) an illegal page fault happens in a RMCB or HW interrupt, (lock all pages!)
+  2) all available pages have been locked,
+  3) the application is using non-committed pages for null pointer protection.
+"Double Fault" - multiple exceptions occurred
+"Invalid TSS" - typically due to RMCB or HW interrupt being called after the
+   selectors/memory have been deallocated (remember to reset the mouse)
+"General Protection Fault" - bad parameter sent to a DPMI call
+
+"80386 required."
+
+Since 80286 and lesser processors don't have the hardware necessary to
+run CWSDPMI.  No workaround, upgrade.
+
+"DOS 3 required."
+
+A few interrupts are used which need DOS 3.0 or higher.  I don't expect to
+ever see this message, since 80386 machines were introduced after DOS 3.0
+and that check is made first.
+
+"CWSDPMI V0.90+ (r7) Copyright (C) 2010 CW Sandmann  ABSOLUTELY NO WARRANTY"
+
+An informational message displayed if the program is not run in one-pass mode.
+
+"Protected mode not accessible."
+
+This message should only be displayed if running CWSDPMI in a protected
+environment with no access to protected mode.  In this case, DPMI should
+already be available and CWSDPMI would not be needed.  This might happen if
+a 16-bit DPMI client is loaded and a DJGPP image attempts to load CWSDPMI
+to provide 32-bit DPMI services under Windows.
+
+"Warning: cannot open swap file c:\cwsdpmi.swp"
+
+Maybe you are out of file handles, or the swap file name is incorrectly
+specified in the image (change the name with cwsparam).
+
+"No swap space!"
+
+This message means you tried to use more swap file than CWSDPMI was
+configured to handle.  Since this is protected against in the memory
+allocation code you should never see this message, except when swapping
+is disabled and you run out of DOS memory for page tables (greater than
+400MB of memory without 4MB pages).
+
+"Swap disk full!"
+
+This means the swap file could not be expanded when trying to swap
+memory out to disk.  This would normally not be seen, unless you are
+writing output to the same disk which holds the swap file.  Decrease
+the amount of memory your DPMI application is using or free up disk space.
+
+"Interrupt 0x??"
+
+Your application tried to call an interrupt from protected mode which
+normally shouldn't be called (something like a data pointer).  If the
+request was allowed to continue it would likely hang your machine.  If you
+see this message and think the interrupt should be allowed to continue, let
+me know.
+
+"Error: Using XMS switched CPU into V86 mode."
+
+This message might be seen if you have your memory manager in AUTO mode.  The
+only workaround in this case is to stop using AUTO mode.
+
+"Error: could not allocate page table memory"
+
+The page table memory (a minimum of 16Kb) is allocated from conventional
+memory (either in the 640Kb region or UMBs).  If CWSDPMI cannot allocate the
+minimum necessary memory, you would see this message.  Free up some
+conventional memory.  You may also see this message if a page directory needs
+to be faulted in, and there are no available pages.  This means too many pages
+have been locked for the allocated page tables available.  While CWSDPMI
+tries to dynamically allocate these if needed, this effort failed.  You need
+to increase the number of page tables with CWSPARAM, or increase the amount
+of free conventional memory if it is low.  If the application which calls
+CWSDPMI internally manages all the DOS memory, the page tables may need to
+be pre-allocated at DPMI startup time (if this is needed, try using the
+run option flag 2 in cwsparam).
+
+"16-bit DPMI unsupported."
+
+CWSDPMI is a 32-bit only DPMI server.  Ideally, on the request to enter DPMI's
+PM with a 16-bit request, we would just fail the call setting the carry bit
+like the DPMI specification describes.  Some buggy 16-bit compiler tools don't
+check the return status and will hang the machine in this case.  So, I issue
+an error message and exit the image instead.
+
+"Descriptors exhausted."
+
+An attempt to nest a DPMI client failed in the setup phase due to insufficient
+free selectors in the LDT.
+
+"CWSDPMI not removed"
+
+When the -u parameter is specified, if DPMI is not detected this message is
+printed.  Informational.
diff --git a/cwsdpmi/bin/cwsdpmi.exe b/cwsdpmi/bin/cwsdpmi.exe
index 2285a74..6865892 100644
--- a/cwsdpmi/bin/cwsdpmi.exe
+++ b/cwsdpmi/bin/cwsdpmi.exe
diff --git a/cwsdpmi/bin/cwsdpr0.exe b/cwsdpmi/bin/cwsdpr0.exe
index 1bbac32..565d2f0 100644
--- a/cwsdpmi/bin/cwsdpr0.exe
+++ b/cwsdpmi/bin/cwsdpr0.exe
diff --git a/cwsdpmi/bin/cwsdstub.exe b/cwsdpmi/bin/cwsdstub.exe
index 5388d00..15399ca 100644
--- a/cwsdpmi/bin/cwsdstub.exe
+++ b/cwsdpmi/bin/cwsdstub.exe
diff --git a/cwsdpmi/bin/cwsparam.doc b/cwsdpmi/bin/cwsparam.doc
index 8ed0f97..7c425bf 100644
--- a/cwsdpmi/bin/cwsparam.doc
+++ b/cwsdpmi/bin/cwsparam.doc
@@ -1,102 +1,103 @@
-CWSPARAM                provided by Charles W Sandmann (sandmann@clio.rice.edu)

-                                    1206 Braelinn, Sugar Land, TX 77479

-CWSPARAM OVERVIEW

-

-There are certain parameters built into the CWSDPMI.EXE, CWSDPR0.EXE, and

-CWSDSTUB.EXE images which might need to be modified for distributing some

-programs for better performance.  The modified image could be renamed, and

-then the DPMI server name in a DJGPP V2 image may be stubedited to use the

-new name.  CWSPARAM can also be run directly on images created by appending

-COFF images to CWSDSTUB.EXE.  The parameters which can be modified are

-discussed below:

-

-Full name of paging file ("" to disable) ? [c:\cwsdpmi.swp]

-

- This is where the file which provides virtual memory should be located.  It

- may be desirable to change the default to a faster disk or one with more

- free space.  To disable paging, enter two double quotes.  The file name

- MUST contain the disk, full path, and file name.  Anything less may cause

- unpredictable behavior.  The maximum length of the name is 48 characters,

- and it must be a DOS 8.3 style name.  Press return to keep the current value.

-

-Number of page tables to initially allocate (0=auto) ? [0]

-

- Page tables are stored in the DOS memory area and consume this precious

- resource.  The default value of 0 chooses automatic allocation of enough

- page tables to hold all of the physical memory without paging (with the

- minimum enough to hold 20Mb without paging and a maximum to hold 128Mb).

- If your program is a fixed size (and does not spawn any other DPMI images)

- you can specify one page table for each 4Mb of your program's size.  If

- your program is small, this can save a small amount of DOS memory, which

- might be needed for other purposes.  If your program is large (typically

- larger than the physical memory) specifying a number will consume more DOS

- memory but improve performance since the page tables themselves will not

- page.  CWSDPMI will attempt to dynamically allocate additional page tables

- at run time to prevent paging, but if all of the DOS memory is free,

- performance would be degraded.  Remember to add an extra page table if

- mapping physical devices.  As of r5, you must also allocate additional

- space for the bitmaps for physical and virtual memory (one bit for each

- 4K page).  Add one page for each 128Mb of total address space

- (physical+virtual).  Be generous, since someone may run the image on a

- 512Mb machine (or larger) and require 4 pages for the physical bitmap

- even if you disable virtual memory.

-

-Minimum application memory desired before 640K paging ? [512Kb]

-

- Typically, CWSDPMI only uses extended memory to return as DPMI memory, and

- saves the DOS memory area for DOS requests.  On small memory machines with

- applications not needing much DOS memory (or poorly tuned memory management)

- there may be very little extended memory left available, with up to 600K of

- DOS memory unused.  This parameter determines the threshold which CWSDPMI

- should start using DOS memory as DPMI memory.  If an application is known to

- not need much DOS memory, this value could be set to the size of the memory

- required for the image, which provides roughly 0.5Mb additional memory before

- paging happens.  This parameter is one I would expect might be adjusted when

- distributing an application for maximum performance on a wide variety of

- machines.

-

-Paragraphs of DOS memory to reserve when 640K paging ? [3840]

-

- If the image needs to page in the DOS memory area (triggered by the parameter

- value above), this parameter tells how much DOS memory to save for DOS

- allocation requests.  The value is in paragraphs (16 byte increments), so the

- default is 60Kbytes.  You will actually end up with a bit more probably, since

- the page tables are allocated on 4Kbyte rounded boundaries.  Some amount of

- DOS memory needs to be saved for potentially expanding the file table, the

- sbrk() algorithm, DMA buffers, etc.  This amount can be tuned on an image

- by image basis, and may be as little as 0.

-

-Paragraphs of memory for extra CWSDPMI internal heap ? [256]

-

- CWSDPMI is built with an internal 4K heap.  Each nested task consumes around

- 300 bytes, each memory zone takes 12 bytes, and each HW interrupt (nested)

- takes around 850 bytes.  If extra DOS memory is available, the additional

- paragraphs specified here are added to the heap at execution time by DOS.

- The 8K default should be good for up to 40Mb in small requests.  If you

- compile very large C++ programs or allocate lots of memory in small pieces,

- you might need to bump this parameter.  If you are very tight on memory,

- you could decrease it to zero and pick up another 4K of DOS memory.  Each

- paragraph allows a little more than one more memory zone (or in the worse

- case around 85Kb of DPMI memory).

-

-Maximum size of swap file ? [128Mb]

-

- CWSDPMI dynamically allocates memory at run time for up to 2044Mb of swap file

- space.  The bitmap for the swap usage requires 32 bytes of memory per 1Mb of

- swap usage.  The default will provide up to 128Mb of swap space with a run

- time memory requirement of 4Kb DOS memory.  The dynamic allocation is limited

- by the free space on the drive, so if you don't have much space on your hard

- drive, you don't need to worry about this.  If you run small programs and want

- to save memory (maybe to stuff the allocation in a UMB) you can decrease this

- value.  If you want to run monster programs needing 2Gb of address space you

- can increase this value (but this would increase the DOS memory usage).  This

- value is ignored by CWSDPR0.

-

-Value of run option flags ? [0]

-

- There are a few run-time options in CWSDPMI which should not be needed by most

- users which can be activated by this field of bit flags:

-   Bit 0 (1):  Disable allocation of run time page tables in UMBs (low only)

-   Bit 1 (2):  Pre-allocate page table memory (in DPMI memory request)

-   Bit 2 (4):  Disable DPMI 1.0 extensions (Null page protection, mapping)

- Other values should be zero for future compatibility but are currently ignored.

+CWSPARAM                provided by Charles W Sandmann (cwsdpmi@earthlink.net)
+                                    1206 Braelinn, Sugar Land, TX 77479
+CWSPARAM OVERVIEW
+
+There are certain parameters built into the CWSDPMI.EXE, CWSDPR0.EXE, and
+CWSDSTUB.EXE images which might need to be modified for distributing some
+programs for better performance.  The modified image could be renamed, and
+then the DPMI server name in a DJGPP V2 image may be stubedited to use the
+new name.  CWSPARAM can also be run directly on images created by appending
+COFF images to CWSDSTUB.EXE.  The parameters which can be modified are
+discussed below:
+
+Full name of paging file ("" to disable) ? [c:\cwsdpmi.swp]
+
+ This is where the file which provides virtual memory should be located.  It
+ may be desirable to change the default to a faster disk or one with more
+ free space.  To disable paging, enter two double quotes.  The file name
+ MUST contain the disk, full path, and file name.  Anything less may cause
+ unpredictable behavior.  The maximum length of the name is 48 characters,
+ and it must be a DOS 8.3 style name.  Press return to keep the current value.
+
+Number of page tables to initially allocate (0=auto) ? [0]
+
+ Page tables are stored in the DOS memory area and consume this precious
+ resource.  The default value of 0 chooses automatic allocation of enough
+ page tables to hold all of the physical memory without paging (with the
+ minimum enough to hold 20Mb without paging and a maximum to hold 128Mb).
+ If your program is a fixed size (and does not spawn any other DPMI images)
+ you can specify one page table for each 4Mb of your program's size.  If
+ your program is small, this can save a small amount of DOS memory, which
+ might be needed for other purposes.  If your program is large (typically
+ larger than the physical memory) specifying a number will consume more DOS
+ memory but improve performance since the page tables themselves will not
+ page.  CWSDPMI will attempt to dynamically allocate additional page tables
+ at run time to prevent paging, but if all of the DOS memory is free,
+ performance would be degraded.  Remember to add an extra page table if
+ mapping physical devices.  As of r5, you must also allocate additional
+ space for the bitmaps for physical and virtual memory (one bit for each
+ 4K page).  Add one page for each 128Mb of total address space
+ (physical+virtual).  Be generous, since someone may run the image on a
+ 512Mb machine (or larger) and require 4 pages for the physical bitmap
+ even if you disable virtual memory.
+
+Minimum application memory desired before 640K paging ? [512Kb]
+
+ Typically, CWSDPMI only uses extended memory to return as DPMI memory, and
+ saves the DOS memory area for DOS requests.  On small memory machines with
+ applications not needing much DOS memory (or poorly tuned memory management)
+ there may be very little extended memory left available, with up to 600K of
+ DOS memory unused.  This parameter determines the threshold which CWSDPMI
+ should start using DOS memory as DPMI memory.  If an application is known to
+ not need much DOS memory, this value could be set to the size of the memory
+ required for the image, which provides roughly 0.5Mb additional memory before
+ paging happens.  This parameter is one I would expect might be adjusted when
+ distributing an application for maximum performance on a wide variety of
+ machines.
+
+Paragraphs of DOS memory to reserve when 640K paging ? [3840]
+
+ If the image needs to page in the DOS memory area (triggered by the parameter
+ value above), this parameter tells how much DOS memory to save for DOS
+ allocation requests.  The value is in paragraphs (16 byte increments), so the
+ default is 60Kbytes.  You will actually end up with a bit more probably, since
+ the page tables are allocated on 4Kbyte rounded boundaries.  Some amount of
+ DOS memory needs to be saved for potentially expanding the file table, the
+ sbrk() algorithm, DMA buffers, etc.  This amount can be tuned on an image
+ by image basis, and may be as little as 0.
+
+Paragraphs of memory for extra CWSDPMI internal heap ? [256]
+
+ CWSDPMI is built with an internal 4K heap.  Each nested task consumes around
+ 300 bytes, each memory zone takes 12 bytes, and each HW interrupt (nested)
+ takes around 850 bytes.  If extra DOS memory is available, the additional
+ paragraphs specified here are added to the heap at execution time by DOS.
+ The 8K default should be good for up to 40Mb in small requests.  If you
+ compile very large C++ programs or allocate lots of memory in small pieces,
+ you might need to bump this parameter.  If you are very tight on memory,
+ you could decrease it to zero and pick up another 4K of DOS memory.  Each
+ paragraph allows a little more than one more memory zone (or in the worse
+ case around 85Kb of DPMI memory).
+
+Maximum size of swap file ? [128Mb]
+
+ CWSDPMI dynamically allocates memory at run time for up to 2044Mb of swap file
+ space.  The bitmap for the swap usage requires 32 bytes of memory per 1Mb of
+ swap usage.  The default will provide up to 128Mb of swap space with a run
+ time memory requirement of 4Kb DOS memory.  The dynamic allocation is limited
+ by the free space on the drive, so if you don't have much space on your hard
+ drive, you don't need to worry about this.  If you run small programs and want
+ to save memory (maybe to stuff the allocation in a UMB) you can decrease this
+ value.  If you want to run monster programs needing 2Gb of address space you
+ can increase this value (but this would increase the DOS memory usage).  This
+ value is ignored by CWSDPR0.
+
+Value of run option flags ? [0]
+
+ There are a few run-time options in CWSDPMI which should not be needed by most
+ users which can be activated by this field of bit flags:
+   Bit 0 (1):  Disable allocation of run time page tables in UMBs (low only)
+   Bit 1 (2):  Pre-allocate page table memory (in DPMI memory request)
+   Bit 2 (4):  Disable DPMI 1.0 extensions (Null page protection, mapping)
+   Bit 3 (8):  Disable use of 4MB pages (very large memory support)
+ Other values should be zero for future compatibility but are currently ignored.
diff --git a/cwsdpmi/bin/mswitch.asm b/cwsdpmi/bin/mswitch.asm
deleted file mode 100644
index 32df5ed..0000000
--- a/cwsdpmi/bin/mswitch.asm
+++ /dev/null
@@ -1,536 +0,0 @@
-; Copyright (C) 1995-2002 CW Sandmann (sandmann@clio.rice.edu) 1206 Braelinn, Sugar Land, TX 77479

-; Copyright (C) 1993 DJ Delorie, 24 Kirsten Ave, Rochester NH 03867-2954

-;

-; This file is distributed under the terms listed in the document

-; "copying.cws", available from CW Sandmann at the address above.

-; A copy of "copying.cws" should accompany this file; if not, a copy

-; should be available from where this file was obtained.  This file

-; may not be distributed without a verbatim copy of "copying.cws".

-;

-; This file is distributed WITHOUT ANY WARRANTY; without even the implied

-; warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

-

-; Modified for VCPI Implement by Y.Shibata Aug 5th 1991

-; Modified for PC98 A20 line control by kaho Aug 5 1998

-

-	title	switch between real and protected mode

-	include segdefs.inc

-	include tss.inc

-	include gdt.inc

-	include vcpi.inc

-

-;------------------------------------------------------------------------

-

-	start_data16

-

-	extrn	_gdt:gdt_s

-	extrn	_gdt_phys:gdt_s

-	extrn	_idt_phys:gdt_s

-	extrn	_tss_ptr:word

-	extrn	_use_xms:byte

-	extrn	_hard_master_lo:byte

-	extrn	_hard_master_hi:byte

-	extrn	_hard_slave_lo:byte

-	extrn	_hard_slave_hi:byte

-	extrn	_mtype:byte

-

-	extrn	_vcpi_installed:byte	;VCPI Installed set this not Zero

-	extrn	_vcpi_entry:fword

-	extrn	_abs_client:dword

-	extrn	_DPMIsp:word

-	extrn	_saved_interrupt_vector:dword

-

-	end_data16

-	start_bss

-

-	public	_dr

-_dr	label	dword

-_dr0	dd	?

-_dr1	dd	?

-_dr2	dd	?

-_dr3	dd	?

-	dd	?

-	dd	?

-_dr6	dd	?

-_dr7	dd	?

-

-	public	_c_tss, _a_tss, _o_tss, _f_tss

-_c_tss	label	tss_s	; for "real mode" state

-	db	type tss_s dup (?)

-_a_tss	label	tss_s	; for running program

-	db	type tss_s dup (?)

-_o_tss	label	tss_s	; for convenience functions

-	db	type tss_s dup (?)

-_f_tss	label	tss_s	; for page handling

-	db	type tss_s dup (?)

-

-	public	_features

-_features	dd	?		; CPUID features

-	public	_was_exception

-_was_exception	db	?		; exceptions set this to 1

-

-	end_bss

-

-;------------------------------------------------------------------------

-

-	start_code16

-

-	extrn	_xms_local_enable_a20:near

-	extrn	_xms_local_disable_a20:near

-

-real_stack	dd	?

-real_idt	dw	03ffh, 0000h, 0000h

-

-	public	_go32

-_go32	proc	near

-	push	si

-	push	di

-	push	bx

-_go32_1:

-;	push	0b000h				;dbg

-;	pop	fs				;dbg

-;	mov	word ptr fs:[0f01h],07h		;dbg - normal

-	mov	ax,DGROUP			; Compute physical offset

-	xor	dx,dx				; of tss_ptr

-	shld	dx,ax,4				; Seg * 16

-	shl	ax,4

-	add	ax,_tss_ptr			; plus tss_ptr

-	adc	dx,0

-	mov	_gdt[g_atss].base0,ax			; Trick here - make atss

-	mov	_gdt[g_atss].base1,dl			; selector point to

-	mov	_gdt[g_atss].base2,dh			; tss_ptr (jmpt g_atss)

-

-

-	mov	al,0fdh					; clear busy flag

-	and	byte ptr _gdt[g_atss].stype,al

-	and	byte ptr _gdt[g_ctss].stype,al		; non-VCPI LDTR below

-

-	mov	_was_exception,0

-

-; set real_stack for return from _go32

-	mov	word ptr cs:real_stack,sp

-	mov	word ptr cs:real_stack+2,ss

-	movzx	esp,sp				; make sure it's OK

-

-	cli

-

-	cmp	_vcpi_installed,0

-	je	short real_to_protect		;Not VCPI Mode

-	mov	esi,_abs_client

-	mov	ax,VCPI_MODE_CHANGE

-	int	VCPI_REQ			;Change Protect Mode

-;end_loop0:

-;	jmp	short end_loop0			;Never Come here!! Safety Loop

-

-real_to_protect:

-	call	_set_a20

-	or	_gdt[g_rdata].lim1,40h		;Set the big bit (non-vcpi only)

-	lgdt	fword ptr _gdt_phys

-	lidt	fword ptr _idt_phys

-

-	mov	eax,cr0

-	or	al,1

-	mov	cr0,eax				; we're in protected mode!

-	db	0eah				; far jmp

-	dw	offset go_protect_far_jump

-	dw	g_rcode

-;

-;	Entry Protect Mode

-;

-	public	_protect_entry

-_protect_entry	label	near

-

-go_protect_far_jump:

-	cli

-	mov     ax,g_rdata

-	mov	ds,ax

-	mov	es,ax

-	mov	fs,ax

-	mov	gs,ax

-	mov	ss,ax

-	movzx	esp,word ptr cs:real_stack	;Need VCPI Inialize

-	xor	eax,eax

-	mov	cr2,eax		 ;zero so we can tell INT 0E from page fault

-

-	mov	eax,_dr0

-	mov	dr0,eax

-	mov	eax,_dr1

-	mov	dr1,eax

-	mov	eax,_dr2

-	mov	dr2,eax

-	mov	eax,_dr3

-	mov	dr3,eax

-	mov	eax,_dr6

-	mov	dr6,eax

-	mov	eax,_dr7

-	mov	dr7,eax

-

-	test	byte ptr _features,8		;Page Size Extensions?

-	jz	short nopse

-	db	0fh,20h,0e0h			;mov eax,cr4

-	or	al,10h				;Enable 4Mb pages

-	db	0fh,22h,0e0h			;mov cr4,eax

-nopse:

-

-	cmp	_vcpi_installed,0

-	jne	short no_tss_load		;Now Paging Mode in VCPI

-	mov	bx,_tss_ptr

-	mov	eax,[bx].tss_cr3

-	or	eax,eax				;Test for zero

-	je	short set_paging_far_jump

-	mov	cr3,eax

-	mov	eax,cr0

-	or	eax,80000000h

-	mov	cr0,eax				; paging enabled!

-	db	0eah				; far jmp

-	dw	offset set_paging_far_jump

-	dw	g_rcode

-set_paging_far_jump:

-	mov	ax,g_ctss

-	ltr	ax

-	jmp	short no_tss_load		; Chip errata fix from Symantec

-no_tss_load:

-	jmpt	g_atss				; load state from VCPU

-

-_go32	endp

-

-; _go_real_mode must follow the task jump, so a task jump to return

-; is valid

-

-	public	_go_real_mode

-_go_real_mode	proc	near

-

-	mov	eax,dr6

-	mov	_dr6,eax

-

-	cmp	_vcpi_installed,0

-	je	short protect_to_real

-

-;Make VCPI call to return to real

-	clts				;TS Clear

-	mov	eax,DGROUP

-	push	eax			;GS

-	push	eax			;FS

-	push	eax			;DS

-	push	eax			;ES

-	push	eax			;SS

-	movzx	eax,word ptr cs:real_stack

-	push	eax			;ESP

-	pushfd				;EFLAGS

-	mov	ax,_TEXT		;high word still zero

-	push	eax			;CS

-	mov	ax,offset back_to_v86	;high word still zero

-	push	eax			;EIP

-;	movzx	esp,sp

-

-	mov	ax,g_core

-	mov	ds,ax

-	mov	ax,VCPI_MODE_CHANGE

-	assume	es:DGROUP

-	call	fword ptr es:_vcpi_entry

-;end_loop1:

-;	jmp	short end_loop1		;Never Come here!! Safety Loop

-

-protect_to_real:

-	and	_gdt[g_rdata].lim1,0bfh		;Clear the big bit (non-vcpi only)

-	push	ss

-	pop	ss				;Reload

-	mov	eax,cr0

-	and	eax,07ffffff6h ; clear PE, TS, PG

-	mov	cr0,eax

-

-	db	0eah				; far jmp

-	dw	offset back_to_real_far_jump

-	dw	_TEXT

-;

-;	Entry Real Mode

-;

-back_to_real_far_jump:

-

-	lidt	fword ptr cs:real_idt

-	lss	sp,cs:real_stack

-back_to_v86:

-

-	mov	ax,ss

-	mov	ds,ax

-	mov	es,ax

-	mov	fs,ax

-	mov	gs,ax

-

-;	call	_reset_a20

-

-	cmp	_was_exception,0

-	je	short not_hard

-

-	mov	bx,_tss_ptr

-	mov	al,[bx].tss_irqn

-

-;	push	0b000h				;dbg

-;	pop	fs				;dbg

-;	mov	ah, 70h				;dbg - reverse

-;	mov	word ptr fs:[0f00h],ax		;dbg

-

-	mov	bx,16				;pseudo IRQ

-	cmp	al,1ch

-	je	short is_hard_1

-	xor	bx,bx				;IRQ base

-	cmp	al,_hard_master_lo

-	jb	short try_slavepic

-	cmp	al,_hard_master_hi

-	jbe	short is_hard

-try_slavepic:

-	mov	bl,8				;IRQ base

-	cmp	al,_hard_slave_lo

-	jb	short not_hard

-	cmp	al,_hard_slave_hi

-	ja	short not_hard

-

-is_hard:

-;	mov	cl,_hard_slave_lo

-;	add	cl,5

-;	cmp	al,cl

-;	je	short not_hard		; for NPX errors

-

-;	mov	cl,_hard_master_lo

-;	inc	cl

-;	cmp	al,cl

-;	je	short not_hard		; to check for ^C

-

-	mov	ah,al

-	and	ah,7			; IRQ offset

-	add	bl,ah			; IRQ #

-is_hard_1:

-	shl	bx,2

-	add	bx,offset _saved_interrupt_vector	;ES is already DS

-	cmp	word ptr [bx+2],0

-	jne	short redirect

-

-	push	0			; Not redirected, use interrupt table

-	pop	es

-	movzx	bx,al

-	shl	bx,2			; 4 bytes per interrupt entry

-redirect:

-	push	_DPMIsp			; Needed for RMCB's to be recursive

-	mov	_DPMIsp,sp

-	sub	_DPMIsp,spare_stack

-	push	3002h

-	call	dword ptr es:[bx]		; Really an interrupt

-	pop	_DPMIsp

-	jmp	_go32_1

-

-not_hard:

-	pop	bx

-	pop	di

-	pop	si

-	ret

-

-_go_real_mode	endp

-

-;------------------------------------------------------------------------

-	public	_reset_a20

-_reset_a20	proc	near

-	cmp	_vcpi_installed,0

-	jne	short reset_a20_nop

-	cmp	_use_xms,0

-	je	short reset_a20_local

-	call	_xms_local_disable_a20

-reset_a20_nop:

-	ret

-

-reset_a20_pc98:

-	mov	al,3

-	out	0f6h,al		; disable the A20 line

-	ret

-reset_a20_local:

-	cmp	_mtype,0	; PC98 raw mode

-	jne	short reset_a20_pc98

-	in	al,092h		; 092h PS/2 & clone system control port "A"

-	and	al,not 2	; this resets the A20 bit in register al

-	jmp	short $+2	; forget the instruction fetch

-	out	092h,al		; set the A20 bit off

-	ret

-_reset_a20	endp

-

-	public	_set_a20

-_set_a20 proc	near

-	cmp	_vcpi_installed,0

-	jne	short set_a20_nop

-	call	check_a20	; If already enabled skip A20 enable

-	jz	short set_a20_nop

-	cmp	_use_xms,0

-	je	short set_a20_local

-	call	_xms_local_enable_a20

-set_a20_nop:

-	ret

-

-set_a20_pc98:

-	mov	al,2

-	out	0f6h,al		; enable the A20 line

-	ret	

-set_a20_local:

-	cmp	_mtype,0	; PC98 raw mode

-	jne	short set_a20_pc98

-	pushf

-	cli

-

-	in	al,092h		; 092h PS/2 & clone system control port "A"

-	or	al,2		; this sets the A20 bit in register al

-	jmp	short $+2	; forget the instruction fetch

-	out	092h,al		; set the A20 bit on

-

-	call	check_a20

-	jz	short a20_done

-

-need_to_set_a20:

-	call	waitkb

-	mov	al,0d1h

-	out	64h,al

-	call	waitkb

-	mov	al,0dfh		; Patrick

-	out	60h,al

-	call	waitkb

-	mov	al,0ffh		; Patrick

-	out	64h,al		; Patrick

-	call	waitkb		; Patrick

-

-wait_for_valid_a20:

-	call	check_a20

-	jnz	short wait_for_valid_a20

-a20_done:

-	popf

-	ret

-

-; interrupts disabled, ax, dx, es destroyed, returns z bit if a20 enabled

-check_a20:

-	; Modified by Prashant TR - don't use fs and gs here before switching 

-	; to protmode.  FS/GS left from TC++ 3.0 can kill us when we come here.

-	push	bx

-	push    ds

-	xor	ax,ax		; zero it

-	mov     ds,ax

-	dec	ax		; ax = 0ffffh

-	mov     es,ax

-	xor	bx,bx

-

-	mov     ax,ds:[bx]      ; word from 0:0 (int 0 interrupt)

-	mov	dx,ax		; save it

-	not	ax		; make a different value

-	xchg    ax,es:[bx+10h]  ; modify ffff:10h to known different value

-	xchg    cx,ds:[bx]      ; get ds:[0] with a write

-	cmp	dx,cx		; compare 0:0 with saved; z bit if same = a20

-	mov     es:[bx+10h],ax  ; restore

-	mov     ds:[bx],dx

-

-	pop     ds

-	pop	bx

-	ret

-

-waitkb:

-	xor	cx,cx		; zero it

-waitkb1:

-	jmp	short $+2

-	jcxz	short $+2

-	in	al,64h

-	test	al,2

-	loopnz	waitkb1

-	ret

-_set_a20 endp

-

-;------------------------------------------------------------------------

-

-	public	_cpumode	; 0=real mode, 1=V86

-_cpumode:

-	smsw	ax

-	and	ax,1

-	ret

-

-;------------------------------------------------------------------------

-; Determination of Cpu type.  EAX, EBX, ECX, EDX destroyed.

-;	0  for 8086/80186

-;	2  for 80286

-;	3  for 386

-;	4  for 486

-;	5  for 586 (Pentium) or better (thanks, Morten!)

-

-cpuid	macro

-	db	0fh,0a2h

-	endm

-

-	public	_cputype	; from Intel 80486 reference manual

-_cputype:

-	xor	cx,cx		; default type is 8086

-	pushf

-	pop	bx

-	and	bh,0fh

-	push	bx

-	popf

-	pushf

-	pop	ax

-	and	ax,0f000h

-	cmp	ax,0f000h

-	jz	short cpu8086

-	or	bh,0f0h

-	push	bx

-	popf

-	pushf

-	pop	ax

-	and	ax,0f000h

-	jz	short cpu286

-

-	mov	dx,sp		; Save old SP

-	and	sp,not 3	; Align it! (we will set AC flag)

-	mov	al,18		; AC flag number

-	call	cpuflipflag

-	mov	sp,dx

-	jnc	short cpu386

-	mov	al,21		; ID flag number

-	call	cpuflipflag

-	jnc	short cpu486

-

-;The cpu supports the CPUID instruction.  Maybe Pentium or late 486?

-	xor	eax,eax

-	cpuid			; ebx:edx:ecx = vendor id; eax = max support

-	or	eax,eax

-	mov	al,5

-	jz	short cpu_pentium

-	xor	eax,eax

-	inc	ax		; call with eax = 1

-	cpuid			; edx=features; ah&0f = family; al=model/step

-	mov	al,ah

-	mov	_features,edx

-cpu_pentium:

-	and	ax,0fh

-	ret	

-cpu486:

-	inc	cx

-cpu386:

-	inc	cx

-cpu286:

-	inc	cx

-	inc	cx

-cpu8086:				; or 186

-	mov	ax,cx

-	ret

-

-cpuflipflag:		; Try to flip flag #Al in Eflags, Return C=1 if possible

-	push	si	; This plus the pushed ret address still stack aligned

-	movzx	esi,al

-	pushfd

-	pop	eax

-	mov	ebx,eax

-	btc	eax,esi

-	push	eax

-	popfd

-	pushfd

-	pop	eax

-	xor	eax,ebx

-	push	ebx

-	popfd

-	bt	eax,esi

-	pop	si

-	ret

-;------------------------------------------------------------------------

-

-	end_code16

-

-	end

diff --git a/cwsdpmi/bin/readme.upd b/cwsdpmi/bin/readme.upd
deleted file mode 100644
index 6d7cdec..0000000
--- a/cwsdpmi/bin/readme.upd
+++ /dev/null
@@ -1,11 +0,0 @@
-This is an update of the r5 distribution for testing purposes.  It is a test

-version!  The copyright banner has an updated year but the images are very

-similar to the standard r5 binaries.  Three bug fixes:

-

-1) Test of PSE feature checks correct bit (removes warnings with BOCHS).

-2) Load ESP properly to clear upper bits with VCPI servers setting them

-   (Martin, please test!)

-3) Serializing jump to avoid Pentium III bug hang (Symantec bug).

-

-MSWITCH.ASM includes all three changes.  The copyright year is also changed

-in control.c - but a single byte change isn't worth including here.

diff --git a/cwsdpmi/manifest/csdpmi5b.mft b/cwsdpmi/manifest/csdpmi5b.mft
deleted file mode 100644
index b8e2f91..0000000
--- a/cwsdpmi/manifest/csdpmi5b.mft
+++ /dev/null
@@ -1,8 +0,0 @@
-manifest/csdpmi5b.mft

-manifest/csdpmi5b.ver

-bin/cwsdpmi.doc

-bin/cwsdpmi.exe

-bin/cwsdpr0.exe

-bin/cwsparam.doc

-bin/cwsparam.exe

-bin/cwsdstub.exe

diff --git a/cwsdpmi/manifest/csdpmi5b.ver b/cwsdpmi/manifest/csdpmi5b.ver
deleted file mode 100644
index 2aebd47..0000000
--- a/cwsdpmi/manifest/csdpmi5b.ver
+++ /dev/null
@@ -1 +0,0 @@
-csdpmi5b CWSDPMI binary distribution (release 5)

diff --git a/readmes/dos b/readmes/dos
index 13b9764..d62c71c 100644
--- a/readmes/dos
+++ b/readmes/dos
@@ -7,8 +7,15 @@ CWSDPMI can be downloaded from:
 
     http://clio.rice.edu/cwsdpmi/
 
-NASM itself is distributed under the terms of the GNU Lesser General
-Public License (LGPL), version 2.1, or at your option, any later
-version.  See the file COPYING.  NASM can be downloaded from:
+              NASM, the Netwide Assembler.
+
+Many many developers all over the net respect NASM for what it is
+- a widespread (thus netwide), portable (thus netwide!), very
+flexible and mature assembler tool with support for many output
+formats (thus netwide!!).
+
+NASM itself is distributed under the terms of the "simplified"
+(2-clause) BSD license.  See the file LICENSE.  NASM can be downloaded
+from:
 
     http://www.nasm.us/