[WIP] The ultimate ros amd64 bringup #361

tkreuzer · 2018-02-05T02:17:55Z

Make ReactOS boot to user mode on x64

This fixes a large number of issues and implements missing stuff.

TODO

Fix resource conflict issues properly

…e x64 version of MiInitMachineDependent()

…t all page table PFNs

…atePebOrTeb() The size is in bytes, not in pages! On x86 we got away with it, since PEB and TEB require only a single page and the 1 passed to MiInsertVadEx() was aligned up to PAGE_SIZE. On x64 this doesn't work, since the size is 2 pages.

…nd FFFFF68000000000 in MiInitSystemMemoryAreas()

…pace()

…mGetPageFileMapping, remove obsolete functions.

…on x64

…port for win32k syscalls in KiSystemCallHandler

… restoring the trap frame

ThFabba · 2018-02-11T08:47:13Z

dll/ntdll/dispatch/amd64/dispatch.S

+    mov r9, [rax + PeKernelCallbackTable]
+
+    ; Call the routine
+    call qword ptr [r9 + r8 * 8]


You need to reserve 32 bytes of stack for it.

No, I dont't. (It's part of the callback frame that has been allocated by the kernel).

Ah, didn't see that. Sounds good.

…erModeIOPL information class, instead just don't do anything.

… duplication

The code uses FAST486 to emulate the BIOS code.

…nted HAL functions.

ThFabba · 2018-02-11T22:00:47Z

hal/halx86/amd64/x86bios.c

@@ -12,7 +12,7 @@
 //#define NDEBUG
 #include <debug.h>

-//#include "x86emu.h"
+#include <fast486.h>


tkreuzer · 2018-02-06T22:34:01Z

hal/halx86/acpi/halacpi.c

@@ -994,8 +994,8 @@ HalpBuildAcpiResourceList(IN PIO_RESOURCE_REQUIREMENTS_LIST ResourceList)

        /* Get the interrupt number */
        Interrupt = HalpPicVectorRedirect[HalpFixedAcpiDescTable.sci_int_vector];
-        ResourceList->List[0].Descriptors[0].u.Interrupt.MinimumVector = Interrupt;
-        ResourceList->List[0].Descriptors[0].u.Interrupt.MaximumVector = Interrupt;
+        ResourceList->List[0].Descriptors[0].u.Interrupt.MinimumVector = Interrupt + PRIMARY_VECTOR_BASE;


It is a hack that I applied, which ensures that at least HAL is loaded properly and doesn't have any resource conflicts. I have doubts that what you claim about untranslated resources is true, since you can see in the registry on Windows that e.g. "HKEY_LOCAL_MACHINE\HARDWARE\RESOURCEMAP\Hardware Abstraction Layer\ACPI x64 platform.Raw" contains all vectors starting from 0 and ".Translated" contains the same vectors.

tkreuzer · 2018-02-06T22:34:43Z

hal/halx86/generic/dma.c

@@ -152,8 +152,8 @@ HalpInitDma(VOID)
        * Check if Extended DMA is available. We're just going to do a random
        * read and write.
        */
-        WRITE_PORT_UCHAR((PUCHAR)FIELD_OFFSET(EISA_CONTROL, DmaController2Pages.Channel2), 0x2A);
-        if (READ_PORT_UCHAR((PUCHAR)FIELD_OFFSET(EISA_CONTROL, DmaController2Pages.Channel2)) == 0x2A)
+        WRITE_PORT_UCHAR((PUCHAR)(ULONG_PTR)FIELD_OFFSET(EISA_CONTROL, DmaController2Pages.Channel2), 0x2A);


Dunno, maybe the base address is 0?

tkreuzer · 2018-02-06T22:40:28Z

ntoskrnl/ke/amd64/stubs.c

+
+    /* Finally switch RSP to the new stack.
+       We pass OldStackBase to make sure it is not lost. */
+    OldStackBase = KiSwitchKernelStackHelper(StackOffset, OldStackBase);


Yes, and if we are honest, we can't do it this way on x86 either, unless all the code that leads to this is entirely written in assembly. For x64 I have a solution that is "by the book", i.e. it returns to asm code, then switches the stack and repeats the system call handling.

tkreuzer · 2018-02-06T22:41:20Z

ntoskrnl/ke/amd64/stubs.c

+    PULONG_PTR NewStack;
+    PULONG_PTR UserStackPointer;
+    NTSTATUS CallbackStatus;
+#ifdef _M_IX86


Yes, that was the plan. This looks generally pretty portable to me.

tkreuzer · 2018-02-06T22:49:29Z

ntoskrnl/ke/amd64/stubs.c

+    /* Check if we have GDI Batch operations */
+    if (GdiBatchCount)
+    {
+        *UserStackPointer -= 256;


Much stack, so large, very structure.

tkreuzer · 2018-02-06T23:41:38Z

ntoskrnl/mm/amd64/init.c

@@ -15,13 +15,13 @@
 #include <debug.h>

 #include <mm/ARM3/miarm.h>
+#include <fltkernel.h>


It contains a few nice macros like Add2Ptr, PtrOffset, ROUND_TO_SIZE, IS_ALIGNED, We should probably put those somewhere separate and include them from fltkernel.h and ntoskrnl.h, so we don't need the whole clusterfuck, but still have those macros available.

tkreuzer · 2018-02-06T23:43:55Z

ntoskrnl/mm/amd64/init.c

-    MiSessionViewStart = (PCHAR)MiSessionViewEnd - MmSessionViewSize;
+    /* Session view is below the session working set */
+    MmSessionViewSize = MI_SESSION_VIEW_SIZE;
+    MiSessionViewEnd = (PUCHAR)MiSessionImageStart - MI_SESSION_WORKING_SET_SIZE;


Not to self: use MiSessionSpaceWs here.

tkreuzer · 2018-02-06T23:47:50Z

ntoskrnl/mm/amd64/init.c

@@ -702,6 +703,10 @@ MiInitMachineDependent(IN PLOADER_PARAMETER_BLOCK LoaderBlock)
    /* Map the PFN database pages */
    MiBuildPfnDatabase(LoaderBlock);

+    /* Initialize the bogus address space */


I probably copy-pasted this comment from i386. It's not "bogus", but it's an address space that was manually initialized. Just don't ask too many questions! It is like it is. Alex made it, so it must be right.

tkreuzer · 2018-02-06T23:49:23Z

ntoskrnl/mm/amd64/page.c

@@ -680,7 +642,7 @@ MmCreateProcessAddressSpace(IN ULONG MinWs,
    PdePte.u.Hard.PageFrameNumber = HyperPfn;
    *SystemPte = PdePte;
    __invlpg(PageTablePointer);
-
+ 


Looks like it is a trailing whitespace.

tkreuzer · 2018-02-06T23:53:57Z

sdk/lib/crt/msvcrtex.cmake

@@ -62,7 +62,6 @@ if(ARCH STREQUAL "i386")
        math/i386/cisqrt.c)
 elseif(ARCH STREQUAL "amd64")
    list(APPEND MSVCRTEX_ASM_SOURCE
-        except/amd64/chkstk_asm.s
        except/amd64/chkstk_ms.s)


Nothing I guess. It might be from the time when GCC handled things differently, but IIRC this is not the case anymore for x86 and I don't care for GCC/x64.

ThFabba · 2018-02-18T08:47:26Z

base/services/dhcpcsvc/dhcp/options.c

@@ -451,7 +451,8 @@ pretty_print_option(unsigned int code, unsigned char *data, int len,
 	static char optbuf[32768]; /* XXX */
 	int hunksize = 0, numhunk = -1, numelem = 0;
 	char fmtbuf[32], *op = optbuf;
-	int i, j, k, opleft = sizeof(optbuf);
+    int i, j, k;


Something's off about the indentation here.

zefklop · 2018-03-03T12:48:07Z

sdk/lib/crt/crt.cmake

@@ -462,7 +462,7 @@ elseif(ARCH STREQUAL "amd64")
        math/amd64/log.S
        math/amd64/log10.S
        math/amd64/pow.S
-        math/amd64/sqrt.S
+        # math/amd64/sqrt.S


What's wrong with it ?

zefklop · 2018-03-03T12:55:00Z

ntoskrnl/mm/ARM3/procsup.c

-#else
-    PointerPde = MiAddressToPde(HYPER_SPACE);
-#endif
+    PointerPde = MiAddressToPde((PVOID)HYPER_SPACE);


Can't we add the cast to the definition of HYPER_SPACE ? What's the point of having it defined as an integer when everybody knows it's an address ?

zefklop · 2018-03-03T13:04:37Z

ntoskrnl/ke/amd64/stubs.c

+
+    /* Adjust the PCR fields */
+    __addgsqword(FIELD_OFFSET(KPCR, NtTib.StackBase), StackOffset);
+    __addgsqword(FIELD_OFFSET(KIPCR, Prcb.RspBase), StackOffset);


I fail to see how, except for those two lines, this function is specific to amd64. Can't we factor out a general x86 version ?

zefklop · 2018-03-03T13:08:23Z

ntoskrnl/mm/ARM3/procsup.c

@@ -1381,7 +1381,7 @@ MmDeleteProcessAddressSpace2(IN PEPROCESS Process)
        MiDecrementShareCount(Pfn1, PageFrameIndex);

        /* Page table is now dead. Bye bye... */
-        ASSERT((Pfn1->u3.e2.ReferenceCount == 0) || (Pfn1->u3.e1.WriteInProgress));
+        ///ASSERT((Pfn1->u3.e2.ReferenceCount == 0) || (Pfn1->u3.e1.WriteInProgress));


Please #ifdef it out, with a relevant comment.

HBelusca · 2018-03-24T00:38:44Z

win32ss/drivers/videoprt/int10.c

@@ -321,7 +349,11 @@ IntInt10CallBios(
                               FALSE,
                               NULL);

+#ifdef _M_AMD64
+    Status = x86BiosCall(0x10, &BiosContext) ? STATUS_SUCCESS : STATUS_UNSUCCESSFUL;;


Double trailing semi-colon.

GeoB99 · 2018-03-25T10:16:15Z

Apologise for my ignorance but I believe when 64-bit support gets released out, only 64-bit software can be installed partly because of the lack of WOW support. Am I correct?

nkrapivin · 2018-03-25T10:26:30Z

Yup, you're right. Just find some XP x64 only apps and here you go.

Mouvedia · 2018-04-08T20:58:45Z

@nkrapivin is there an issue for the Windows on Windows support?

ROCKNROLLKID · 2018-07-26T20:24:20Z

So what are we doing with this PR?

HBelusca · 2018-07-26T23:02:57Z

@ROCKNROLLKID : @tkreuzer is slowly reintegrating it in the master.

tkreuzer · 2018-08-19T15:33:10Z

Closing for now.

tkreuzer added 24 commits February 5, 2018 01:53

[NTOS:IO] Fix parsing of resource lists

91c8699

[NTOS:MM] Make sure to call MmInitializeProcessAddressSpace() from th…

ff1eaad

…e x64 version of MiInitMachineDependent()

[NTOS:MM] In the x64 version of MmCreateProcessAddressSpace() zero ou…

15a08e8

…t all page table PFNs

[NTOS:MM] On x64 reserve the address range between FFFF800000000000 a…

286b2fb

…nd FFFFF68000000000 in MiInitSystemMemoryAreas()

[HAL] Fix some 64 bit warnings

12002d4

[NTOS:OB] Fix some 64 bit warnings

546c05b

[CRT] Use chkstk_ms.s on x64 builds

1bc15af

[CRT] Remove the x64 asm version of sqrt from build

1f6d61b

[NTOS:IO] Fix a warning on MSVC builds

9e097de

[NTOS:MM] SLIST handling for kernel stacks

aedb9f6

[NTOS:MM] Handle PPEs und PXEs as well in MmInitializeProcessAddressS…

d139743

…pace()

[NTOS:MM] Simplify and fix x64 version of MiGetPteForProcess(), fix M…

25064d0

…mGetPageFileMapping, remove obsolete functions.

[NTOS:MM] Fix session space initialization on x64

75e11df

[NTOS:MM] Fix paged pool initialization on x64

13ef07a

[NTOS:MM] Fix paged pool expansion

03920b1

[NTOS:SE] Fix SeSetSecurityDescriptorInfoEx to avoid pool corruption …

235566d

…on x64

[CMAKE] Add baseaddress_msvc_x64.cmake

1e326fb

[CSRSRV] Fix type of ViewSize parameter passed to NtMapViewOfSection

c4fbc15

[NTOS:KE] Implement KiConvertToGuiThread, KeSwitchKernelStack and sup…

29db059

…port for win32k syscalls in KiSystemCallHandler

[NTOS:KE] Save and restore previous mode in KiZwSystemService and fix…

c24203c

… restoring the trap frame

[NTOS:KE] Implement KiGetUserModeStackAddress() and KeUserModeCallback()

c9d1f6a

[HAL][NTOS][I8042PRT] Add some hacks related to resource conflicts

e503f53

[NTOS:MM] HACK: comment out an ASSERT that still fails on x64

9335f8f

tkreuzer added the WIP label Feb 5, 2018

tkreuzer self-assigned this Feb 5, 2018

tkreuzer requested review from zefklop, HeisSpiter, ionescu007 and yagoulas February 5, 2018 02:17

ThFabba reviewed Feb 11, 2018

View reviewed changes

tkreuzer added 12 commits February 11, 2018 11:08

[NTDLL] Delete obsolete file amd64/stubs.c

c3380ef

[NTOS:PS] On x64 don't fail in NtSetInformationProcess with ProcessUs…

c249c15

…erModeIOPL information class, instead just don't do anything.

[WIN32K] Fix ULONG/SIZE_T issue

8b48097

[NTOS:KE] Silence KiRundownThread, it has nothing to do

e7e7e40

[VIDEOPRT] Don't try to use NtVdmControl on x64

a9fc91d

[VIDEOPRT] [FORMATTING] No code change

6e0a3be

[VIDEOPRT] Call IntInt10CallBios from VideoPortInt10, instead of code…

f062e2d

… duplication

[VIDEOPRT] Fixes for x64

21177bb

[NDK] Add BIOS call API for amd64

ebe9bc4

[FAST486] DWORD -> ULONG, so it can be used in kernel mode

c8bd63f

[HAL] Implement amd64 BIOS call support

4c80d83

The code uses FAST486 to emulate the BIOS code.

[VIDEOPRT] Implement support for INT10 on x64 using the newly impleme…

96f3020

…nted HAL functions.

ThFabba reviewed Feb 11, 2018

View reviewed changes

tkreuzer commented Feb 15, 2018

View reviewed changes

ThFabba reviewed Feb 18, 2018

View reviewed changes

EricKohl removed their request for review February 19, 2018 19:32

zefklop reviewed Mar 3, 2018

View reviewed changes

HBelusca reviewed Mar 24, 2018

View reviewed changes

sanchaez added enhancement For PRs with an enhancement/new feature. review pending For PRs undergoing review. labels Apr 7, 2018

tkreuzer closed this Aug 19, 2018

binarymaster removed the review pending For PRs undergoing review. label Jun 11, 2023

binarymaster added this to the x64 bringup milestone Jun 11, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] The ultimate ros amd64 bringup #361

[WIP] The ultimate ros amd64 bringup #361

tkreuzer commented Feb 5, 2018 •

edited

ThFabba Feb 11, 2018

tkreuzer Feb 16, 2018

ThFabba Feb 18, 2018

ThFabba Feb 11, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

tkreuzer Feb 6, 2018

ThFabba Feb 18, 2018

zefklop Mar 3, 2018

zefklop Mar 3, 2018

zefklop Mar 3, 2018

zefklop Mar 3, 2018

HBelusca Mar 24, 2018

GeoB99 commented Mar 25, 2018

nkrapivin commented Mar 25, 2018

Mouvedia commented Apr 8, 2018

ROCKNROLLKID commented Jul 26, 2018

HBelusca commented Jul 26, 2018

tkreuzer commented Aug 19, 2018

[WIP] The ultimate ros amd64 bringup #361

[WIP] The ultimate ros amd64 bringup #361

Conversation

tkreuzer commented Feb 5, 2018 • edited

Make ReactOS boot to user mode on x64

TODO

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

GeoB99 commented Mar 25, 2018

nkrapivin commented Mar 25, 2018

Mouvedia commented Apr 8, 2018

ROCKNROLLKID commented Jul 26, 2018

HBelusca commented Jul 26, 2018

tkreuzer commented Aug 19, 2018

tkreuzer commented Feb 5, 2018 •

edited