- Ros-dev - ReactOS.org

Patch - somebody please review (FPU state saving)

by Gregor Anich

Hi! Here's a patch which mainly adds FPU state saving support on taskswitching. It does not protect User/Kernel mode "mixup" because I want this to be discussed further before it is implemented. I have also attached a program which tests if the FPU state is saved, it creates 2 threads which both do some FPU calculations and checks the result. One thing which is changed in the patch is that the taskswitching function (Ki386ContextSwitch) does not calculate the TSS->Esp0 from NewThread->InitialStack but saves it on the stack and restores it. For a v86 call there seems to be no more reason to change the InitialStack of the thread (InitialStack seems to be used only for setting up TSS->Esp0 on taskswitch - not anymore - and for IoGetInitialStack) so a v86 call saves and changes TSS->Esp0 instead of Thread->InitialStack. I have been talking with Filip about having 2 FPU save areas, one for kernel and one for user-mode... he was against it because of the stack space which is taken away. He also told me that the stack size cannot be changed because some drivers rely on it to calculate the remaining stack. What we could do is 1) Use KeSave/RestoreFloatingState() in KM when FPU is used. 2) Compile kernel-mode code with soft-fpu switch to make the compiler call a softfpu lib instead of using the x87 3) Put the FPU save areas somewhere else. Maybe we could allocate MM_STACK_SIZE + 2 * sizeof (FX_SAVE_AREA) and set InitialStack to StackBase + MM_STACK_SIZE (like now) so the stack size available for code would not change. It would also be cool to save the state only when needed. Maybe this could be done with a per-processor pointer to a FX_SAVE_AREA which is set to the last thread's FX_SAVE_AREA which used the FPU. When a not-present exception happens (after a taskswitch where TS in %cr0 was set) we check if the pointer is NULL, if it is not we save the current state into the area it points to and set it to point to the current thread's FX_SAVE_AREA and load the saved state (if any) of the current thread. When a thread terminates and per-processor FX_SAVE_AREA pointer points to the terminated thread's FX_SAVE_AREA we set it to NULL. Thanks, blight Index: include/ntos/tss.h =================================================================== RCS file: /CVS/ReactOS/reactos/include/ntos/tss.h,v retrieving revision 1.4 diff -u -r1.4 tss.h --- include/ntos/tss.h 5 Jan 2004 14:28:19 -0000 1.4 +++ include/ntos/tss.h 29 Oct 2004 14:55:49 -0000 @@ -5,7 +5,8 @@ #ifndef __INCLUDE_DDK_I386_TSS_H #define __INCLUDE_DDK_I386_TSS_H -#define KTSS_ESP0 (0x4) +#define KTSS_ESP0 (0x4) +#define KTSS_EFLAGS (0x24) #define KTSS_IOMAPBASE (0x66) #ifndef __ASM__ Index: ntoskrnl/include/internal/trap.h =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/include/internal/trap.h,v retrieving revision 1.4 diff -u -r1.4 trap.h --- ntoskrnl/include/internal/trap.h 8 Sep 2002 10:23:23 -0000 1.4 +++ ntoskrnl/include/internal/trap.h 3 Nov 2004 17:43:44 -0000 @@ -27,10 +27,9 @@ #ifndef __NTOSKRNL_INCLUDE_INTERNAL_TRAP_H #define __NTOSKRNL_INCLUDE_INTERNAL_TRAP_H -#define TF_SAVED_ORIG_STACK (0x8C) -#define TF_REGS (0x90) -#define TF_ORIG_EBP (0x94) - +#define TF_SAVED_EXCEPTION_STACK (0x8C) +#define TF_REGS (0x90) +#define TF_ORIG_EBP (0x94) #ifndef __ASM__ @@ -40,7 +39,7 @@ { KTRAP_FRAME Tf; - ULONG SavedInitialStack; + ULONG SavedExceptionStack; /* * These are put on the top of the stack by the routine that entered Index: ntoskrnl/include/internal/i386/fpu.h =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/include/internal/i386/fpu.h,v retrieving revision 1.5 diff -u -r1.5 fpu.h --- ntoskrnl/include/internal/i386/fpu.h 12 Oct 2004 00:33:30 -0000 1.5 +++ ntoskrnl/include/internal/i386/fpu.h 22 Oct 2004 22:37:26 -0000 @@ -20,11 +20,20 @@ #ifndef __NTOSKRNL_INCLUDE_INTERNAL_I386_FPU_H #define __NTOSKRNL_INCLUDE_INTERNAL_I386_FPU_H +#define SIZEOF_FX_SAVE_AREA 528 + +#ifndef __ASM__ + +#include <internal/i386/ke.h> + extern ULONG HardwareMathSupport; VOID KiCheckFPU(VOID); +NTSTATUS +KiHandleFpuFault(PKTRAP_FRAME Tf, ULONG ExceptionNr); + typedef struct _FNSAVE_FORMAT { ULONG ControlWord; ULONG StatusWord; @@ -62,4 +71,7 @@ ULONG Cr0NpxState; } FX_SAVE_AREA, *PFX_SAVE_AREA; +#endif /* !__ASM__ */ + #endif /* __NTOSKRNL_INCLUDE_INTERNAL_I386_FPU_H */ + Index: ntoskrnl/include/internal/i386/ke.h =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/include/internal/i386/ke.h,v retrieving revision 1.15 diff -u -r1.15 ke.h --- ntoskrnl/include/internal/i386/ke.h 31 Oct 2004 12:49:36 -0000 1.15 +++ ntoskrnl/include/internal/i386/ke.h 4 Nov 2004 00:17:41 -0000 @@ -69,9 +69,17 @@ #define KTRAP_FRAME_RESERVED9 (0x8A) #define KTRAP_FRAME_SIZE (0x8C) +#define X86_EFLAGS_IF 0x00000200 /* Interrupt Enable flag */ +#define X86_EFLAGS_IOPL 0x00003000 /* I/O Privilege Level bits */ +#define X86_EFLAGS_NT 0x00004000 /* Nested Task flag */ #define X86_EFLAGS_VM 0x00020000 /* Virtual Mode */ #define X86_EFLAGS_ID 0x00200000 /* CPUID detection flag */ +#define X86_CR0_NE 0x00000020 /* enable native FPU error reporting */ +#define X86_CR0_TS 0x00000008 /* enable exception on FPU instruction for task switch */ +#define X86_CR0_EM 0x00000004 /* enable FPU emulation (disable FPU) */ +#define X86_CR0_MP 0x00000002 /* enable FPU monitoring */ + #define X86_CR4_PAE 0x00000020 /* enable physical address extensions */ #define X86_CR4_PGE 0x00000080 /* enable global pages */ @@ -85,6 +93,11 @@ #define X86_EXT_FEATURE_3DNOW 0x40000000 /* 3DNOW! extension present */ +/* Possible values for KTHREAD's NpxState */ +#define NPX_STATE_INVALID 0x01 +#define NPX_STATE_VALID 0x02 +#define NPX_STATE_DIRTY 0x04 + #ifndef __ASM__ typedef struct _KTRAP_FRAME @@ -187,12 +200,20 @@ : "r" (X)); #define Ke386SaveFlags(x) __asm__ __volatile__("pushfl ; popl %0":"=g" (x): /* no input */) #define Ke386RestoreFlags(x) __asm__ __volatile__("pushl %0 ; popfl": /* no output */ :"g" (x):"memory") -#define Ke386GetCr4() ({ \ + +#define _Ke386GetCr(N) ({ \ unsigned int __d; \ - __asm__("movl %%cr4,%0\n\t" :"=r" (__d)); \ + __asm__("movl %%cr" #N ",%0\n\t" :"=r" (__d)); \ __d; \ }) -#define Ke386SetCr4(X) __asm__ __volatile__("movl %0,%%cr4": :"r" (X)); +#define _Ke386SetCr(N,X) __asm__ __volatile__("movl %0,%%cr" #N : :"r" (X)); + +#define Ke386GetCr0() _Ke386GetCr(0) +#define Ke386SetCr0(X) _Ke386SetCr(0,X) +#define Ke386GetCr2() _Ke386GetCr(2) +#define Ke386SetCr2(X) _Ke386SetCr(2,X) +#define Ke386GetCr4() _Ke386GetCr(4) +#define Ke386SetCr4(X) _Ke386SetCr(4,X) static inline void Ki386Cpuid(ULONG Op, PULONG Eax, PULONG Ebx, PULONG Ecx, PULONG Edx) { @@ -207,6 +228,7 @@ #elif defined(_MSC_VER) + #define Ke386DisableInterrupts() __asm cli #define Ke386EnableInterrupts() __asm sti #define Ke386HaltProcessor() __asm hlt Index: ntoskrnl/include/internal/i386/ps.h =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/include/internal/i386/ps.h,v retrieving revision 1.16 diff -u -r1.16 ps.h --- ntoskrnl/include/internal/i386/ps.h 30 Oct 2004 23:48:56 -0000 1.16 +++ ntoskrnl/include/internal/i386/ps.h 3 Nov 2004 19:02:51 -0000 @@ -26,6 +26,7 @@ #define KTHREAD_STACK_LIMIT 0x1C #define KTHREAD_TEB 0x20 #define KTHREAD_KERNEL_STACK 0x28 +#define KTHREAD_NPX_STATE 0x31 #define KTHREAD_APCSTATE_PROCESS 0x44 #define KTHREAD_SERVICE_TABLE 0xDC #define KTHREAD_PREVIOUS_MODE 0x137 @@ -51,12 +52,12 @@ #pragma pack(push,4) -// Fixme: Use correct types? +/* Fixme: Use correct types? */ typedef struct _KPROCESSOR_STATE { PCONTEXT ContextFrame; PVOID SpecialRegisters; } KPROCESSOR_STATE; - + /* ProcessoR Control Block */ typedef struct _KPRCB { USHORT MinorVersion; Index: ntoskrnl/ke/main.c =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/main.c,v retrieving revision 1.202 diff -u -r1.202 main.c --- ntoskrnl/ke/main.c 24 Oct 2004 15:26:14 -0000 1.202 +++ ntoskrnl/ke/main.c 3 Nov 2004 17:43:51 -0000 @@ -297,6 +297,7 @@ ASSERT(FIELD_OFFSET(KTHREAD, InitialStack) == KTHREAD_INITIAL_STACK); ASSERT(FIELD_OFFSET(KTHREAD, Teb) == KTHREAD_TEB); ASSERT(FIELD_OFFSET(KTHREAD, KernelStack) == KTHREAD_KERNEL_STACK); + ASSERT(FIELD_OFFSET(KTHREAD, NpxState) == KTHREAD_NPX_STATE); ASSERT(FIELD_OFFSET(KTHREAD, ServiceTable) == KTHREAD_SERVICE_TABLE); ASSERT(FIELD_OFFSET(KTHREAD, PreviousMode) == KTHREAD_PREVIOUS_MODE); ASSERT(FIELD_OFFSET(KTHREAD, TrapFrame) == KTHREAD_TRAP_FRAME); @@ -307,6 +308,7 @@ ASSERT(FIELD_OFFSET(KPROCESS, IopmOffset) == KPROCESS_IOPM_OFFSET); ASSERT(FIELD_OFFSET(KPROCESS, LdtDescriptor) == KPROCESS_LDT_DESCRIPTOR0); ASSERT(FIELD_OFFSET(KTRAP_FRAME, Reserved9) == KTRAP_FRAME_RESERVED9); + ASSERT(FIELD_OFFSET(KV86M_TRAP_FRAME, SavedExceptionStack) == TF_SAVED_EXCEPTION_STACK); ASSERT(FIELD_OFFSET(KV86M_TRAP_FRAME, regs) == TF_REGS); ASSERT(FIELD_OFFSET(KV86M_TRAP_FRAME, orig_ebp) == TF_ORIG_EBP); @@ -314,6 +316,12 @@ ASSERT(FIELD_OFFSET(KPCR, Self) == KPCR_SELF); ASSERT(FIELD_OFFSET(KPCR, PrcbData) + FIELD_OFFSET(KPRCB, CurrentThread) == KPCR_CURRENT_THREAD); + ASSERT(FIELD_OFFSET(KTSS, Esp0) == KTSS_ESP0); + ASSERT(FIELD_OFFSET(KTSS, Eflags) == KTSS_EFLAGS); + ASSERT(FIELD_OFFSET(KTSS, IoMapBase) == KTSS_IOMAPBASE); + + ASSERT(sizeof(FX_SAVE_AREA) == SIZEOF_FX_SAVE_AREA); + LdrInit1(); KeLowerIrql(DISPATCH_LEVEL); Index: ntoskrnl/ke/i386/bthread.S =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/bthread.S,v retrieving revision 1.10 diff -u -r1.10 bthread.S --- ntoskrnl/ke/i386/bthread.S 31 Oct 2004 19:46:10 -0000 1.10 +++ ntoskrnl/ke/i386/bthread.S 3 Nov 2004 19:05:17 -0000 @@ -141,18 +141,6 @@ addl $24, %esp .L3: - /* - * Load the floating point registers - */ - movl (_HardwareMathSupport), %eax - testl %eax,%eax - jz .L2 - testl $(CONTEXT_FLOATING_POINT & ~CONTEXT_i386), %ebx - jz .L2 - frstor (%esp) -.L2: - addl $112, %esp - /* Load the rest of the thread's user mode context. */ movl $0, %eax jmp KeReturnFromSystemCallWithHook Index: ntoskrnl/ke/i386/exp.c =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/exp.c,v retrieving revision 1.93 diff -u -r1.93 exp.c --- ntoskrnl/ke/i386/exp.c 22 Oct 2004 20:32:48 -0000 1.93 +++ ntoskrnl/ke/i386/exp.c 23 Oct 2004 18:26:36 -0000 @@ -40,6 +40,10 @@ #define _STR(x) #x #define STR(x) _STR(x) +#ifndef ARRAY_SIZE +# define ARRAY_SIZE(x) (sizeof (x) / sizeof (x[0])) +#endif + extern void interrupt_handler2e(void); extern void interrupt_handler2d(void); @@ -84,6 +88,7 @@ "Stack Segment Fault", "General Protection", "Page Fault", + "Reserved(15)", "Math Fault", "Alignment Check", "Machine Check" @@ -106,6 +111,7 @@ STATUS_STACK_OVERFLOW, STATUS_ACCESS_VIOLATION, STATUS_ACCESS_VIOLATION, + STATUS_ACCESS_VIOLATION, /* RESERVED */ STATUS_ACCESS_VIOLATION, /* STATUS_FLT_INVALID_OPERATION */ STATUS_DATATYPE_MISALIGNMENT, STATUS_ACCESS_VIOLATION @@ -167,7 +173,7 @@ } else { - if (ExceptionNr < 16) + if (ExceptionNr < ARRAY_SIZE(ExceptionToNtStatus)) { Er.ExceptionCode = ExceptionToNtStatus[ExceptionNr]; } @@ -190,7 +196,7 @@ ULONG KiDoubleFaultHandler(VOID) { - unsigned int cr2_; + unsigned int cr2; ULONG StackLimit; ULONG StackBase; ULONG Esp0; @@ -211,15 +217,7 @@ Esp0 = OldTss->Esp; /* Get CR2 */ -#if defined(__GNUC__) - __asm__("movl %%cr2,%0\n\t" : "=d" (cr2_)); -#elif defined(_MSC_VER) - __asm mov eax, cr2; - __asm mov cr2_, eax; -#else -#error Unknown compiler for inline assembler -#endif - + cr2 = Ke386GetCr2(); if (PsGetCurrentThread() != NULL && PsGetCurrentThread()->ThreadsProcess != NULL) { @@ -245,7 +243,7 @@ /* * Print out the CPU registers */ - if (ExceptionNr < 19) + if (ExceptionNr < ARRAY_SIZE(ExceptionTypeStrings)) { DbgPrint("%s Exception: %d(%x)\n", ExceptionTypeStrings[ExceptionNr], ExceptionNr, 0); @@ -257,7 +255,7 @@ DbgPrint("CS:EIP %x:%x ", OldTss->Cs, OldTss->Eip); KeRosPrintAddress((PVOID)OldTss->Eip); DbgPrint("\n"); - DbgPrint("cr2 %x cr3 %x ", cr2_, OldCr3); + DbgPrint("cr2 %x cr3 %x ", cr2, OldCr3); DbgPrint("Proc: %x ",PsGetCurrentProcess()); if (PsGetCurrentProcess() != NULL) { @@ -420,7 +418,7 @@ /* * Print out the CPU registers */ - if (ExceptionNr < 19) + if (ExceptionNr < ARRAY_SIZE(ExceptionTypeStrings)) { DbgPrint("%s Exception: %d(%x)\n", ExceptionTypeStrings[ExceptionNr], ExceptionNr, Tf->ErrorCode&0xffff); @@ -529,7 +527,7 @@ * Complete CPU context */ { - unsigned int cr2_; + unsigned int cr2; NTSTATUS Status; ULONG Esp0; @@ -540,22 +538,15 @@ Esp0 = (ULONG)&Tf->Eip; /* Get CR2 */ -#if defined(__GNUC__) - __asm__("movl %%cr2,%0\n\t" : "=d" (cr2_)); -#elif defined(_MSC_VER) - __asm mov eax, cr2; - __asm mov cr2_, eax; -#else -#error Unknown compiler for inline assembler -#endif - Tf->DebugPointer = (PVOID)cr2_; + cr2 = Ke386GetCr2(); + Tf->DebugPointer = (PVOID)cr2; /* * If this was a V86 mode exception then handle it specially */ if (Tf->Eflags & (1 << 17)) { - return(KeV86Exception(ExceptionNr, Tf, cr2_)); + return(KeV86Exception(ExceptionNr, Tf, cr2)); } /* @@ -574,7 +565,7 @@ */ if (ExceptionNr == 14) { - if (Ke386NoExecute && Tf->ErrorCode & 0x10 && cr2_ >= KERNEL_BASE) + if (Ke386NoExecute && Tf->ErrorCode & 0x10 && cr2 >= KERNEL_BASE) { KEBUGCHECKWITHTF(ATTEMPTED_EXECUTE_OF_NOEXECUTE_MEMORY, 0, 0, 0, 0, Tf); } @@ -585,7 +576,7 @@ Status = MmPageFault(Tf->Cs&0xffff, &Tf->Eip, &Tf->Eax, - cr2_, + cr2, Tf->ErrorCode); if (NT_SUCCESS(Status)) { @@ -606,15 +597,27 @@ } /* + * Try to handle device-not-present and math-fault exceptions. + */ + if (ExceptionNr == 7 || ExceptionNr == 16) + { + Status = KiHandleFpuFault(Tf, ExceptionNr); + if (NT_SUCCESS(Status)) + { + return(0); + } + } + + /* * Handle user exceptions differently */ if ((Tf->Cs & 0xFFFF) == USER_CS) { - return(KiUserTrapHandler(Tf, ExceptionNr, (PVOID)cr2_)); + return(KiUserTrapHandler(Tf, ExceptionNr, (PVOID)cr2)); } else { - return(KiKernelTrapHandler(Tf, ExceptionNr, (PVOID)cr2_)); + return(KiKernelTrapHandler(Tf, ExceptionNr, (PVOID)cr2)); } } Index: ntoskrnl/ke/i386/fpu.c =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/fpu.c,v retrieving revision 1.13 diff -u -r1.13 fpu.c --- ntoskrnl/ke/i386/fpu.c 15 Aug 2004 16:39:05 -0000 1.13 +++ ntoskrnl/ke/i386/fpu.c 4 Nov 2004 02:48:39 -0000 @@ -32,6 +32,33 @@ #define NDEBUG #include <internal/debug.h> +/* DEFINES *******************************************************************/ +#define EXCEPTION_FLT_DENORMAL_OPERAND (0xc000008dL) +#define EXCEPTION_FLT_DIVIDE_BY_ZERO (0xc000008eL) +#define EXCEPTION_FLT_INEXACT_RESULT (0xc000008fL) +#define EXCEPTION_FLT_INVALID_OPERATION (0xc0000090L) +#define EXCEPTION_FLT_OVERFLOW (0xc0000091L) +#define EXCEPTION_FLT_STACK_CHECK (0xc0000092L) +#define EXCEPTION_FLT_UNDERFLOW (0xc0000093L) + +/* x87 Status Word exception flags */ +#define X87_SW_IE (1<<0) /* Invalid Operation */ +#define X87_SW_DE (1<<1) /* Denormalized Operand */ +#define X87_SW_ZE (1<<2) /* Zero Devide */ +#define X87_SW_OE (1<<3) /* Overflow */ +#define X87_SW_UE (1<<4) /* Underflow */ +#define X87_SW_PE (1<<5) /* Precision */ +#define X87_SW_SE (1<<6) /* Stack Fault */ + +#define X87_SW_ES (1<<7) /* Error Summary */ + +#if 0 +#define X87_SW_C0 (1<<8) /* Condition code */ +#define X87_SW_C1 (1<<9) /* Condition code */ +#define X87_SW_C2 (1<<10) /* Condition code */ +#define X87_SW_C3 (1<<14) /* Condition code */ +#endif + /* GLOBALS *******************************************************************/ ULONG HardwareMathSupport; @@ -42,56 +69,41 @@ KiCheckFPU(VOID) { unsigned short int status; - int cr0_; + int cr0; HardwareMathSupport = 0; - -#if defined(__GNUC__) - __asm__("movl %%cr0, %0\n\t" : "=a" (cr0_)); - /* Set NE and MP. */ - cr0_ = cr0_ | 0x22; - /* Clear EM */ - cr0_ = cr0_ & (~0x4); - __asm__("movl %0, %%cr0\n\t" : : "a" (cr0_)); + cr0 = Ke386GetCr0(); + cr0 |= X86_CR0_NE | X86_CR0_MP; + cr0 &= ~X86_CR0_EM; + Ke386SetCr0(cr0); +#if defined(__GNUC__) __asm__("clts\n\t"); __asm__("fninit\n\t"); __asm__("fstsw %0\n\t" : "=a" (status)); - if (status != 0) - { - __asm__("movl %%cr0, %0\n\t" : "=a" (cr0_)); - /* Set the EM flag in CR0 so any FPU instructions cause a trap. */ - cr0_ = cr0_ | 0x4; - __asm__("movl %0, %%cr0\n\t" : - : "a" (cr0_)); - return; - } - /* fsetpm for i287, ignored by i387 */ - __asm__(".byte 0xDB, 0xE4\n\t"); #elif defined(_MSC_VER) - __asm mov eax, cr0; - __asm mov cr0_, eax; - cr0_ |= 0x22; /* Set NE and MP. */ - cr0_ &= ~0x4; /* Clear EM */ __asm { - mov eax, cr0_; - mov cr0, eax; clts; fninit; fstsw status } +#else +#error Unknown compiler for inline assembler +#endif + if (status != 0) { - __asm mov eax, cr0_; - __asm or eax, 4; /* Set the EM flag in CR0 so any FPU instructions cause a trap. */ - __asm mov cr0, eax; + /* Set the EM flag in CR0 so any FPU instructions cause a trap. */ + Ke386SetCr0(Ke386GetCr0() | X86_CR0_EM); return; } + /* fsetpm for i287, ignored by i387 */ +#if defined(__GNUC__) + __asm__(".byte 0xDB, 0xE4\n\t"); +#elif defined(_MSC_VER) __asm _emit 0xDB __asm _emit 0xe4 -// __asm fsetpm; -// __asm__(".byte 0xDB, 0xE4\n\t"); #else #error Unknown compiler for inline assembler #endif @@ -109,6 +121,12 @@ { char *FpState; + /* check if we are doing software emulation */ + if (!HardwareMathSupport) + { + return STATUS_ILLEGAL_FLOAT_CONTEXT; + } + FpState = ExAllocatePool(PagedPool, FPU_STATE_SIZE); if (NULL == FpState) { @@ -146,3 +164,104 @@ return STATUS_SUCCESS; } + +NTSTATUS +KiHandleFpuFault(PKTRAP_FRAME Tf, ULONG ExceptionNr) +{ + if (ExceptionNr == 7) /* device not present */ + { + unsigned int cr0 = Ke386GetCr0(); + PFX_SAVE_AREA FxSaveArea = (PFX_SAVE_AREA)((char *)KeGetCurrentThread()->InitialStack - sizeof (FX_SAVE_AREA)); + + ASSERT((cr0 & X86_CR0_TS) == X86_CR0_TS); + ASSERT((Tf->Eflags & X86_EFLAGS_VM) == 0); + ASSERT((cr0 & X86_CR0_EM) == 0); + + Ke386DisableInterrupts(); + + cr0 &= ~X86_CR0_TS; + Ke386SetCr0(cr0); + + DPRINT("Device not present exception happened! (Cr0 = 0x%x, NpxState = 0x%x)\n", Ke386GetCr0(), KeGetCurrentThread()->NpxState); + if (KeGetCurrentThread()->NpxState & NPX_STATE_VALID) + { +#if 1 + DPRINT("Restoring previous FPU state\n"); + DPRINT("ControlWord = 0x%08x\n", FxSaveArea->U.FnArea.ControlWord); + DPRINT("StatusWord = 0x%08x\n", FxSaveArea->U.FnArea.StatusWord); + DPRINT("TagWord = 0x%08x\n", FxSaveArea->U.FnArea.TagWord); + DPRINT("ErrorOffset = 0x%08x\n", FxSaveArea->U.FnArea.ErrorOffset); + DPRINT("ErrorSelector = 0x%08x\n", FxSaveArea->U.FnArea.ErrorSelector); + DPRINT("DataOffset = 0x%08x\n", FxSaveArea->U.FnArea.DataOffset); + DPRINT("DataSelector = 0x%08x\n", FxSaveArea->U.FnArea.DataSelector); + DPRINT("Thread->InitialStack = 0x%08x\n", (ULONG)KeGetCurrentThread()->InitialStack); + DPRINT("Thread->KernelStack = 0x%08x\n", (ULONG)KeGetCurrentThread()->KernelStack); +#endif + __asm__ __volatile__("frstor %0" : : "m"(*FxSaveArea)); + } + else /* NpxState & NPX_STATE_INVALID */ + { + DPRINT("Setting up clean FPU state\n"); + __asm__ __volatile__("finit"); + } + KeGetCurrentThread()->NpxState |= NPX_STATE_DIRTY; + DPRINT("Device not present exception handled!\n"); + Ke386EnableInterrupts(); + return STATUS_SUCCESS; + } + else + { + WORD FpuStatusWord; + EXCEPTION_RECORD Er; + CONTEXT Context; + KPROCESSOR_MODE PreviousMode; + + ASSERT(ExceptionNr == 16); /* math fault */ + + PreviousMode = ((Tf->Cs & 0xffff) == USER_CS) ? (UserMode) : (KernelMode); + DPRINT("Math-fault happened! (PreviousMode = %s)\n", + (PreviousMode == UserMode) ? ("UserMode") : ("KernelMode")); + /* Get fpu status and clear exception flags */ + __asm__("fnstenv %0" : "=m"(Context.FloatSave)); + __asm__("fnclex"); /* FIXME: clear only handled exceptions? */ + FpuStatusWord = Context.FloatSave.StatusWord & 0xffff; + + Ke386EnableInterrupts(); + + DPRINT("FpuStatusWord = 0x%04x\n", FpuStatusWord); + + /* Dispatch exception */ + if (FpuStatusWord & X87_SW_IE) + Er.ExceptionCode = EXCEPTION_FLT_INVALID_OPERATION; + else if (FpuStatusWord & X87_SW_DE) + Er.ExceptionCode = EXCEPTION_FLT_DENORMAL_OPERAND; + else if (FpuStatusWord & X87_SW_ZE) + Er.ExceptionCode = EXCEPTION_FLT_DIVIDE_BY_ZERO; + else if (FpuStatusWord & X87_SW_OE) + Er.ExceptionCode = EXCEPTION_FLT_OVERFLOW; + else if (FpuStatusWord & X87_SW_UE) + Er.ExceptionCode = EXCEPTION_FLT_UNDERFLOW; + else if (FpuStatusWord & X87_SW_PE) + Er.ExceptionCode = EXCEPTION_FLT_INEXACT_RESULT; + else if (FpuStatusWord & X87_SW_SE) + Er.ExceptionCode = EXCEPTION_FLT_STACK_CHECK; + else + ASSERT(0); /* not reached */ + + Er.ExceptionFlags = 0; + Er.ExceptionRecord = NULL; + /* FIXME: is this the right way to get the correct EIP of the faulting instruction? */ + Er.ExceptionAddress = (PVOID)Context.FloatSave.ErrorOffset; + Er.NumberParameters = 0; + + /* FIXME: we should pass in a context with floating save area to KiDispatchException */ + DPRINT("Dispatching exception\n"); + KiDispatchException(&Er, NULL, Tf, PreviousMode, TRUE); + + DPRINT("Math-fault handled!\n"); + return STATUS_SUCCESS; + } + + return STATUS_UNSUCCESSFUL; +} + Index: ntoskrnl/ke/i386/multiboot.S =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/multiboot.S,v retrieving revision 1.28 diff -u -r1.28 multiboot.S --- ntoskrnl/ke/i386/multiboot.S 31 Oct 2004 19:46:10 -0000 1.28 +++ ntoskrnl/ke/i386/multiboot.S 3 Nov 2004 19:02:53 -0000 @@ -3,6 +3,7 @@ #include <internal/i386/segment.h> #include <internal/ps.h> #include <internal/i386/mm.h> +#include <internal/i386/fpu.h> #define MULTIBOOT_HEADER_MAGIC (0x1BADB002) @@ -279,7 +280,7 @@ /* * Load the initial kernel stack */ - movl $_init_stack_top, %esp + movl $(_init_stack_top - SIZEOF_FX_SAVE_AREA), %esp /* * Initialize EFLAGS Index: ntoskrnl/ke/i386/thread.c =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/thread.c,v retrieving revision 1.33 diff -u -r1.33 thread.c --- ntoskrnl/ke/i386/thread.c 15 Aug 2004 16:39:05 -0000 1.33 +++ ntoskrnl/ke/i386/thread.c 4 Nov 2004 00:18:55 -0000 @@ -31,13 +31,6 @@ #define NDEBUG #include <internal/debug.h> -/* GLOBALS *******************************************************************/ - -#define FLAG_NT (1<<14) -#define FLAG_VM (1<<17) -#define FLAG_IF (1<<9) -#define FLAG_IOPL ((1<<12)+(1<<13)) - /* FUNCTIONS *****************************************************************/ NTSTATUS @@ -77,10 +70,10 @@ { return(STATUS_UNSUCCESSFUL); } - if ((Context->EFlags & FLAG_IOPL) != 0 || - (Context->EFlags & FLAG_NT) || - (Context->EFlags & FLAG_VM) || - (!(Context->EFlags & FLAG_IF))) + if ((Context->EFlags & X86_EFLAGS_IOPL) != 0 || + (Context->EFlags & X86_EFLAGS_NT) || + (Context->EFlags & X86_EFLAGS_VM) || + (!(Context->EFlags & X86_EFLAGS_IF))) { return(STATUS_UNSUCCESSFUL); } @@ -93,40 +86,37 @@ PULONG KernelStack; ULONG InitSize; PKTRAP_FRAME TrapFrame; + PFX_SAVE_AREA FxSaveArea; /* * Setup a stack frame for exit from the task switching routine */ - InitSize = 5 * sizeof(DWORD) + sizeof(DWORD) + 6 * sizeof(DWORD) + - sizeof(FLOATING_SAVE_AREA) + sizeof(KTRAP_FRAME); + InitSize = 6 * sizeof(DWORD) + sizeof(DWORD) + 6 * sizeof(DWORD) + + + sizeof(KTRAP_FRAME) + sizeof (FX_SAVE_AREA); KernelStack = (PULONG)((char*)Thread->KernelStack - InitSize); /* Set up the initial frame for the return from the dispatcher. */ - KernelStack[0] = 0; /* EDI */ - KernelStack[1] = 0; /* ESI */ - KernelStack[2] = 0; /* EBX */ - KernelStack[3] = 0; /* EBP */ - KernelStack[4] = (ULONG)&PsBeginThreadWithContextInternal; /* EIP */ + KernelStack[0] = (ULONG)Thread->InitialStack - sizeof(FX_SAVE_AREA); /* TSS->Esp0 */ + KernelStack[1] = 0; /* EDI */ + KernelStack[2] = 0; /* ESI */ + KernelStack[3] = 0; /* EBX */ + KernelStack[4] = 0; /* EBP */ + KernelStack[5] = (ULONG)&PsBeginThreadWithContextInternal; /* EIP */ /* Save the context flags. */ - KernelStack[5] = Context->ContextFlags; + KernelStack[6] = Context->ContextFlags; /* Set up the initial values of the debugging registers. */ - KernelStack[6] = Context->Dr0; - KernelStack[7] = Context->Dr1; - KernelStack[8] = Context->Dr2; - KernelStack[9] = Context->Dr3; - KernelStack[10] = Context->Dr6; - KernelStack[11] = Context->Dr7; - - /* Set up the initial floating point state. */ - memcpy((PVOID)&KernelStack[12], (PVOID)&Context->FloatSave, - sizeof(FLOATING_SAVE_AREA)); + KernelStack[7] = Context->Dr0; + KernelStack[8] = Context->Dr1; + KernelStack[9] = Context->Dr2; + KernelStack[10] = Context->Dr3; + KernelStack[11] = Context->Dr6; + KernelStack[12] = Context->Dr7; /* Set up a trap frame from the context. */ - TrapFrame = (PKTRAP_FRAME) - ((char*)KernelStack + 12 * sizeof(DWORD) + sizeof(FLOATING_SAVE_AREA)); + TrapFrame = (PKTRAP_FRAME)(&KernelStack[13]); TrapFrame->DebugEbp = (PVOID)Context->Ebp; TrapFrame->DebugEip = (PVOID)Context->Eip; TrapFrame->DebugArgMark = 0; @@ -149,12 +139,39 @@ TrapFrame->ErrorCode = 0; TrapFrame->Cs = Context->SegCs; TrapFrame->Eip = Context->Eip; - TrapFrame->Eflags = Context->EFlags | FLAG_IF; - TrapFrame->Eflags &= ~(FLAG_VM | FLAG_NT | FLAG_IOPL); + TrapFrame->Eflags = Context->EFlags | X86_EFLAGS_IF; + TrapFrame->Eflags &= ~(X86_EFLAGS_VM | X86_EFLAGS_NT | X86_EFLAGS_IOPL); TrapFrame->Esp = Context->Esp; TrapFrame->Ss = (USHORT)Context->SegSs; /* FIXME: Should check for a v86 mode context here. */ + /* Set up the initial floating point state. */ + if (Context->ContextFlags & CONTEXT_FLOATING_POINT) + { + FxSaveArea = (PFX_SAVE_AREA)((ULONG_PTR)KernelStack + InitSize - sizeof(FX_SAVE_AREA)); + memcpy((PVOID)&FxSaveArea->U.FnArea, (PVOID)&Context->FloatSave, sizeof(FxSaveArea->U.FnArea)); + FxSaveArea->Cr0NpxState = Context->FloatSave.Cr0NpxState; + /* FIXME: for SSE we have to convert FLOATING_SAVE_AREA to FX_SAVE_AREA */ + +#ifndef NDEBUG + DPRINT1("Thread's initial FPU status (0x%x):\n", (ULONG)FxSaveArea); + DPRINT1("ControlWord = 0x%08x\n", FxSaveArea->U.FnArea.ControlWord); + DPRINT1("StatusWord = 0x%08x\n", FxSaveArea->U.FnArea.StatusWord); + DPRINT1("TagWord = 0x%08x\n", FxSaveArea->U.FnArea.TagWord); + DPRINT1("ErrorOffset = 0x%08x\n", FxSaveArea->U.FnArea.ErrorOffset); + DPRINT1("ErrorSelector = 0x%08x\n", FxSaveArea->U.FnArea.ErrorSelector); + DPRINT1("DataOffset = 0x%08x\n", FxSaveArea->U.FnArea.DataOffset); + DPRINT1("DataSelector = 0x%08x\n", FxSaveArea->U.FnArea.DataSelector); + DPRINT1("Thread->InitialStack = 0x%08x\n", (ULONG)Thread->InitialStack); + DPRINT1("Thread->KernelStack = 0x%08x\n", (ULONG)Thread->KernelStack); +#endif + Thread->NpxState = NPX_STATE_VALID; + } + else + { + Thread->NpxState = NPX_STATE_INVALID; + } + /* Save back the new value of the kernel stack. */ Thread->KernelStack = (PVOID)KernelStack; @@ -175,21 +192,25 @@ * Setup a stack frame for exit from the task switching routine */ - KernelStack = (PULONG)((char*)Thread->KernelStack - (8*4)); - KernelStack[0] = 0; /* EDI */ - KernelStack[1] = 0; /* ESI */ - KernelStack[2] = 0; /* EBX */ - KernelStack[3] = 0; /* EBP */ - KernelStack[4] = (ULONG)&PsBeginThread; /* EIP */ - KernelStack[5] = 0; /* Return EIP */ - KernelStack[6] = (ULONG)StartRoutine; /* First argument to PsBeginThread */ - KernelStack[7] = (ULONG)StartContext; /* Second argument to PsBeginThread */ + KernelStack = (PULONG)((char*)Thread->KernelStack - (9 * sizeof(DWORD)) - sizeof(FX_SAVE_AREA)); + KernelStack[0] = (ULONG)Thread->InitialStack - sizeof(FX_SAVE_AREA); /* TSS->Esp0 */ + KernelStack[1] = 0; /* EDI */ + KernelStack[2] = 0; /* ESI */ + KernelStack[3] = 0; /* EBX */ + KernelStack[4] = 0; /* EBP */ + KernelStack[5] = (ULONG)&PsBeginThread; /* EIP */ + KernelStack[6] = 0; /* Return EIP */ + KernelStack[7] = (ULONG)StartRoutine; /* First argument to PsBeginThread */ + KernelStack[8] = (ULONG)StartContext; /* Second argument to PsBeginThread */ Thread->KernelStack = (VOID*)KernelStack; + /* + * Setup FPU state + */ + Thread->NpxState = NPX_STATE_INVALID; + return(STATUS_SUCCESS); } - - - /* EOF */ + Index: ntoskrnl/ke/i386/tskswitch.S =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/tskswitch.S,v retrieving revision 1.19 diff -u -r1.19 tskswitch.S --- ntoskrnl/ke/i386/tskswitch.S 31 Oct 2004 19:46:10 -0000 1.19 +++ ntoskrnl/ke/i386/tskswitch.S 4 Nov 2004 00:32:57 -0000 @@ -17,16 +17,18 @@ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ /* - * FILE: ntoskrnl/ke/i386/tskswitch.S - * PURPOSE: Microkernel thread support + * FILE: ntoskrnl/ke/i386/tskswitch.S + * PURPOSE: Microkernel thread support * PROGRAMMER: David Welch (welch(a)cwcom.net) * UPDATE HISTORY: - * Created 09/10/00 + * Created 09/10/00 */ /* INCLUDES ******************************************************************/ #include <internal/i386/segment.h> +#include <internal/i386/ke.h> +#include <internal/i386/fpu.h> #include <internal/ps.h> #include <ntos/tss.h> #include <internal/ntoskrnl.h> @@ -34,12 +36,13 @@ /* FUNCTIONS ****************************************************************/ .globl _Ki386ContextSwitch +.func Ki386ContextSwitch _Ki386ContextSwitch: -/* +/* * FUNCTIONS: Switches to another thread's context * ARGUMENTS: - * Thread = Thread to switch to - * OldThread = Thread to switch from + * Thread = Thread to switch to + * OldThread = Thread to switch from */ pushl %ebp movl %esp, %ebp @@ -60,7 +63,7 @@ * Get the pointer to the new thread. */ movl 8(%ebp), %ebx - + /* * Set the base of the TEB selector to the base of the TEB for * this thread. @@ -89,7 +92,7 @@ xorl %eax, %eax movl KTHREAD_APCSTATE_PROCESS(%ebx), %edi testw $0xFFFF, KPROCESS_LDT_DESCRIPTOR0(%edi) - jz .L4 + jz 0f pushl KPROCESS_LDT_DESCRIPTOR1(%edi) pushl KPROCESS_LDT_DESCRIPTOR0(%edi) @@ -99,35 +102,46 @@ movl $LDT_SELECTOR, %eax -.L4: +0: lldtw %ax + movl 12(%ebp), %ebx + /* - * Load up the iomap offset for this thread in - * preparation for setting it below. + * FIXME: Save debugging state. */ - movl KPROCESS_IOPM_OFFSET(%edi), %eax /* - * FIXME: Save debugging state. + * Save floating point state if TS was unset. */ + movl %cr0, %eax + testl $X86_CR0_TS, %eax + jnz 1f + movl KTHREAD_INITIAL_STACK(%ebx), %eax + fnsave -SIZEOF_FX_SAVE_AREA(%eax) + movb $NPX_STATE_VALID, KTHREAD_NPX_STATE(%ebx) +1: /* - * FIXME: Save floating point state. + * Load up the iomap offset for this thread in + * preparation for setting it below. */ + movl KPROCESS_IOPM_OFFSET(%edi), %eax + + /* + * Save the stack pointer in this processors TSS + */ + movl %fs:KPCR_TSS, %esi + pushl KTSS_ESP0(%esi) /* * Switch stacks */ - movl 12(%ebp), %ebx movl %esp, KTHREAD_KERNEL_STACK(%ebx) movl 8(%ebp), %ebx movl KTHREAD_KERNEL_STACK(%ebx), %esp movl KTHREAD_STACK_LIMIT(%ebx), %edi - /* - * Set the stack pointer in this processors TSS - */ movl %fs:KPCR_TSS, %esi /* @@ -135,9 +149,6 @@ */ movw %ax, KTSS_IOMAPBASE(%esi) - movl KTHREAD_INITIAL_STACK(%ebx), %eax - movl %eax, KTSS_ESP0(%esi) - /* * Change the address space */ @@ -146,8 +157,16 @@ movl %eax, %cr3 /* - * FIXME: Restore floating point state + * Restore the stack pointer in this processors TSS + */ + popl KTSS_ESP0(%esi) + + /* + * Set TS in cr0 to catch FPU code and load the FPU state when needed */ + movl %cr0, %eax + orl $X86_CR0_TS, %eax + movl %eax, %cr0 /* * FIXME: Restore debugging state @@ -162,9 +181,9 @@ call _KeReleaseSpinLockFromDpcLevel@4 cmpl $0, _PiNrThreadsAwaitingReaping - je .L3 + je 4f call _PiWakeupReaperThread@0 -.L3: +4: /* * Restore the saved register and exit @@ -175,3 +194,5 @@ popl %ebp ret +.endfunc + Index: ntoskrnl/ke/i386/tss.c =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/tss.c,v retrieving revision 1.9 diff -u -r1.9 tss.c --- ntoskrnl/ke/i386/tss.c 15 Aug 2004 16:39:05 -0000 1.9 +++ ntoskrnl/ke/i386/tss.c 27 Oct 2004 19:53:02 -0000 @@ -134,7 +134,7 @@ KeGetCurrentKPCR()->TSS = Tss; /* Initialize the boot TSS. */ - Tss->Esp0 = (ULONG)Ki386InitialStackArray[Id]; + Tss->Esp0 = (ULONG)Ki386InitialStackArray[Id] + MM_STACK_SIZE; /* FIXME: - sizeof(FX_SAVE_AREA)? */ Tss->Ss0 = KERNEL_DS; Tss->IoMapBase = 0xFFFF; /* No i/o bitmap */ Tss->IoBitmap[8192] = 0xFF; @@ -154,9 +154,9 @@ /* Initialize the TSS used for handling double faults. */ TrapTss->Eflags = 0; - TrapTss->Esp0 = ((ULONG)TrapStack + MM_STACK_SIZE); + TrapTss->Esp0 = ((ULONG)TrapStack + MM_STACK_SIZE); /* FIXME: - sizeof(FX_SAVE_AREA)? */ TrapTss->Ss0 = KERNEL_DS; - TrapTss->Esp = ((ULONG)TrapStack + MM_STACK_SIZE); + TrapTss->Esp = ((ULONG)TrapStack + MM_STACK_SIZE); /* FIXME: - sizeof(FX_SAVE_AREA)? */ TrapTss->Cs = KERNEL_CS; TrapTss->Eip = (ULONG)KiTrap8; TrapTss->Ss = KERNEL_DS; @@ -211,7 +211,7 @@ Ki386InitialStackArray[0] = (PVOID)&init_stack; /* Initialize the boot TSS. */ - KiBootTss.Esp0 = (ULONG)&init_stack_top; + KiBootTss.Esp0 = (ULONG)&init_stack_top - sizeof(FX_SAVE_AREA); KiBootTss.Ss0 = KERNEL_DS; // KiBootTss.IoMapBase = FIELD_OFFSET(KTSS, IoBitmap); KiBootTss.IoMapBase = 0xFFFF; /* No i/o bitmap */ @@ -232,9 +232,9 @@ /* Initialize the TSS used for handling double faults. */ KiBootTrapTss.Eflags = 0; - KiBootTrapTss.Esp0 = (ULONG)&trap_stack_top; + KiBootTrapTss.Esp0 = (ULONG)&trap_stack_top; /* FIXME: - sizeof(FX_SAVE_AREA)? */ KiBootTrapTss.Ss0 = KERNEL_DS; - KiBootTrapTss.Esp = (ULONG)&trap_stack_top; + KiBootTrapTss.Esp = (ULONG)&trap_stack_top; /* FIXME: - sizeof(FX_SAVE_AREA)? */ KiBootTrapTss.Cs = KERNEL_CS; KiBootTrapTss.Eip = (ULONG)KiTrap8; KiBootTrapTss.Ss = KERNEL_DS; Index: ntoskrnl/ke/i386/v86m_sup.S =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ke/i386/v86m_sup.S,v retrieving revision 1.11 diff -u -r1.11 v86m_sup.S --- ntoskrnl/ke/i386/v86m_sup.S 31 Oct 2004 19:46:10 -0000 1.11 +++ ntoskrnl/ke/i386/v86m_sup.S 3 Nov 2004 19:12:48 -0000 @@ -71,27 +71,24 @@ * Since we are going to fiddle with the stack pointer this must be * a critical section for this processor */ - - /* - * Save the old initial stack - */ - movl %fs:KPCR_CURRENT_THREAD, %esi - movl KTHREAD_INITIAL_STACK(%esi), %edi - pushl %edi - - /* - * We also need to set the stack in the kthread structure - */ - movl %esp, KTHREAD_INITIAL_STACK(%esi) + cli + + /* + * Save the exception handler stack from the TSS + */ + movl %fs:KPCR_TSS, %esi + pushl KTSS_ESP0(%esi) /* * The stack used for handling exceptions from v86 mode in this thread * will be the current stack adjusted so we don't overwrite the * existing stack frames */ - movl %fs:KPCR_TSS, %esi movl %esp, KTSS_ESP0(%esi) + /* Exit the critical section */ + sti + /* * Create the stack frame for an iret to v86 mode */ @@ -180,18 +177,10 @@ cli /* - * Restore the initial stack + * Restore the exception handler stack in the TSS */ - popl %eax movl %fs:KPCR_TSS, %esi - movl %eax, KTSS_ESP0(%esi) - - /* - * We also need to set the stack in the kthread structure - */ - movl %fs:KPCR_CURRENT_THREAD, %esi - movl KTHREAD_INITIAL_STACK(%esi), %edi - movl %eax, KTHREAD_INITIAL_STACK(%esi) + popl KTSS_ESP0(%esi) /* Exit the critical section */ sti @@ -207,3 +196,4 @@ movl %ebp, %esp popl %ebp ret + Index: ntoskrnl/ps/w32call.c =================================================================== RCS file: /CVS/ReactOS/reactos/ntoskrnl/ps/w32call.c,v retrieving revision 1.17 diff -u -r1.17 w32call.c --- ntoskrnl/ps/w32call.c 30 Oct 2004 23:48:57 -0000 1.17 +++ ntoskrnl/ps/w32call.c 3 Nov 2004 19:13:38 -0000 @@ -44,6 +44,8 @@ PNTSTATUS CallbackStatus; PKTRAP_FRAME SavedTrapFrame; PVOID SavedCallbackStack; + PVOID SavedExceptionStack; + UCHAR SavedNpxState; } NTW32CALL_SAVED_STATE, *PNTW32CALL_SAVED_STATE; typedef struct @@ -81,6 +83,8 @@ PNTW32CALL_SAVED_STATE State; PKTRAP_FRAME SavedTrapFrame; PVOID SavedCallbackStack; + PVOID SavedExceptionStack; + UCHAR SavedNpxState; Thread = PsGetCurrentThread(); if (Thread->Tcb.CallbackStack == NULL) @@ -102,6 +106,8 @@ StackLimit = State->SavedStackLimit; SavedTrapFrame = State->SavedTrapFrame; SavedCallbackStack = State->SavedCallbackStack; + SavedExceptionStack = State->SavedExceptionStack; + SavedNpxState = State->SavedNpxState; /* * Copy the callback status and the callback result to NtW32Call @@ -129,7 +135,9 @@ Thread->Tcb.StackLimit = StackLimit; Thread->Tcb.TrapFrame = SavedTrapFrame; Thread->Tcb.CallbackStack = SavedCallbackStack; - KeGetCurrentKPCR()->TSS->Esp0 = (ULONG)Thread->Tcb.InitialStack; + Thread->Tcb.NpxState = SavedNpxState; + KeGetCurrentKPCR()->TSS->Esp0 = (ULONG)SavedExceptionStack; + Ke386SetCr0(Ke386GetCr0() | X86_CR0_TS); /* set TS */ KeStackSwitchAndRet((PVOID)(OldStack + 1)); /* Should never return. */ @@ -282,9 +290,9 @@ } /* FIXME: Need to check whether we were interrupted from v86 mode. */ - memcpy((char*)NewStack + StackSize - sizeof(KTRAP_FRAME), Thread->Tcb.TrapFrame, - sizeof(KTRAP_FRAME) - (4 * sizeof(DWORD))); - NewFrame = (PKTRAP_FRAME)((char*)NewStack + StackSize - sizeof(KTRAP_FRAME)); + memcpy((char*)NewStack + StackSize - sizeof(KTRAP_FRAME) - sizeof(FX_SAVE_AREA), + Thread->Tcb.TrapFrame, sizeof(KTRAP_FRAME) - (4 * sizeof(DWORD))); + NewFrame = (PKTRAP_FRAME)((char*)NewStack + StackSize - sizeof(KTRAP_FRAME) - sizeof(FX_SAVE_AREA)); NewFrame->Esp -= (ArgumentLength + (4 * sizeof(ULONG))); NewFrame->Eip = (ULONG)LdrpGetSystemDllCallbackDispatcher(); UserEsp = (PULONG)NewFrame->Esp; @@ -304,10 +312,19 @@ SavedState.CallbackStatus = &CallbackStatus; SavedState.SavedTrapFrame = Thread->Tcb.TrapFrame; SavedState.SavedCallbackStack = Thread->Tcb.CallbackStack; + SavedState.SavedExceptionStack = (PVOID)KeGetCurrentKPCR()->TSS->Esp0; + SavedState.SavedNpxState = Thread->Tcb.NpxState; + if (Thread->Tcb.NpxState & NPX_STATE_DIRTY) /* thread has used FPU, save context */ + { + __asm__("fsave %0" : : "m"(*((char *)Thread->Tcb.InitialStack - sizeof(FX_SAVE_AREA)))); + SavedState.SavedNpxState = NPX_STATE_VALID; + Ke386SetCr0(Ke386GetCr0() | X86_CR0_TS); + } Thread->Tcb.InitialStack = Thread->Tcb.StackBase = (char*)NewStack + StackSize; - Thread->Tcb.StackLimit = (ULONG_PTR)NewStack; - Thread->Tcb.KernelStack = (char*)NewStack + StackSize - sizeof(KTRAP_FRAME); - KeGetCurrentKPCR()->TSS->Esp0 = (ULONG)Thread->Tcb.InitialStack; + Thread->Tcb.StackLimit = (ULONG)NewStack; + Thread->Tcb.KernelStack = (char*)NewStack + StackSize - sizeof(KTRAP_FRAME) - sizeof(FX_SAVE_AREA); + Thread->Tcb.NpxState = NPX_STATE_INVALID; + KeGetCurrentKPCR()->TSS->Esp0 = (ULONG)Thread->Tcb.InitialStack - sizeof(FX_SAVE_AREA); KePushAndStackSwitchAndSysRet((ULONG)&SavedState, Thread->Tcb.KernelStack); /*

20 years, 6 months

2
2
0 0

Re: [MinGW-dvlpr] Problem with binutils-2.15.91-20040904-1.tar.gz

by Steven Edwards

--- "Aaron W. LaFramboise" <aaron77thyme(a)aaronwl.com> wrote: > I know nothing about ReactOS internals, but theres a chance someone > might be able to help if you could check all objects with the bad > binutils against all objects with the good binutils, and somewhere > post > the objects that differed. I'd take a look if you could post this > much > information. According to Filip Navara there was a bug in binutils that is fixed in current CVS. I dont really have the time to check all the objects. Thanks Steven __________________________________ Do you Yahoo!? Check out the new Yahoo! Front Page. www.yahoo.com

20 years, 6 months

1
0
0 0

Re: Web site maintenance.

by Lucio Diaz

> > Today's Topics: > > 1. Re: What I have learned on my time off > (LinuxWorld) (Jason Filby) ---------------------------------------------------------------------- > > Message: 1 > Date: Sat, 6 Nov 2004 10:07:57 +0200 > From: Jason Filby <jason.filby(a)gmail.com> > Subject: Re: [ros-dev] What I have learned on my > time off (LinuxWorld) > To: ReactOS Development List <ros-dev(a)reactos.com> > Message-ID: > <e7a1591e04110600072d5b1306(a)mail.gmail.com> > Content-Type: text/plain; charset=US-ASCII > > Steven > > >1. We need a roadmap for the future. > > Agreed. > > >2. We have to have regression testing in place. > > Yes - Casper's doing a lot of great work on this; > with regards to > WineHQ tests, well, we'll have to see if we can't > find a way to > integrate them. Right now let's just focus on > getting the Continuous > Integration System going. > > >3. We need to improve our documentation and > website. > > I've already thought about this a lot; definitely a > wiki for the > Library - with access only to devs to make changes; > a wiki doesn't > have to allow just anyone to make a change and I > don't think that's a > good idea. I've already introduced MedaWiki at work > (it powers > Wikipedia.com) and would really want to get this on > reactos.com for > our Library -- probably a few other pages as well. > > I don't want to see a wiki only site. A wiki doesn't > enforce > structure, which is nice for something like a blog > or list of > compatible hardware/software, or forum. We'll use > the best software > suited to the job -- this means that a lot of effort > will have to go > into integration: search but more importantly login. > > It's just that to get there is a ton of work! The > website needs its > own roadmap, for sure. > Mr. Filby, i have ofered twice to colaborate with my limited capabilities in the proyect, i have offered to actualize the website from time to time, and the made of a FAQ (this will be more difficult, as i will have to test and ask many howtos on reactos) if you find this could be usefull. I am still waiting for an answer. Best regards, Lucio Diaz-Flores. ______________________________________________ Renovamos el Correo Yahoo!: ¡100 MB GRATIS! Nuevos servicios, más seguridad http://correo.yahoo.es

20 years, 6 months

4
3
0 0

Problem with binutils-2.15.91-20040904-1.tar.gz

by Steven Edwards

Hello, The most recent binutils package might have a bug in it but I am not quite sure. When running ReactOS after building with the newer binutils you will see errors like in this thread http://reactos.com:8080/archives/public/ros-dev/2004-November/000492.html Sorry I dont have time to dig in to it further. Thanks Steven __________________________________ Do you Yahoo!? Check out the new Yahoo! Front Page. www.yahoo.com

20 years, 6 months

2
1
0 0

Rosapps broken in current CVS (CTM)

by Gge

The current Rosapps module in CVS fails to build I am using the recommended compiler indicated in Reactos.com Regards Gerard ------------------------------------------------------ mingw32-make[1]: Leaving directory `D:/Gerard/Reactos/Ros/rosapps/sysutils' mingw32-make -C sysutils\ctm mingw32-make[1]: Entering directory `D:/Gerard/Reactos/Ros/rosapps/sysutils/ctm' windres --include-dir ../../../reactos/include --include-dir ../../../reactos/w32api/include ctm.rc -o ctm.coff gcc -I. -I../../../reactos/include -I../../../reactos/w32api/include -pipe -march=i486 -D_M_IX86 -g -c ctm.c -o ctm.o ctm.c:41:20: epsapi.h: No such file or directory ctm.c: In function `PerfDataRefresh': ctm.c:430: warning: assignment makes pointer from integer without a cast ctm.c:433: warning: assignment makes pointer from integer without a cast ctm.c:444: warning: assignment makes pointer from integer without a cast ctm.c:532: warning: assignment makes pointer from integer without a cast mingw32-make[1]: *** [ctm.o] Error 1 mingw32-make[1]: Leaving directory `D:/Gerard/Reactos/Ros/rosapps/sysutils/ctm' mingw32-make: *** [sysutils\ctm] Error 2 D:\Gerard\Reactos\Ros\rosapps> -------------------------------------------------------

20 years, 6 months

1
0
0 0

Crash on current CVS

by Steven Edwards

Hi, I am kind of stuck at the momement as the current CVS wont boot for me. My build env does not quite match the recommended spec for Mingw under Windows so I guess that could be the problem. Is anyone else seeing this? (ke/i386/kernel.c:330) CPU supports PAE mode (ke/i386/kernel.c:341) CPU doesn't run in PAE mode (ke/i386/kernel.c:346) CPU Vendor: AuthenticAMD (ke/i386/kernel.c:350) CPU Model: AMD Athlon(tm) XP 2400+ (ke/i386/kernel.c:353) Ke386CacheAlignment: 64 (ke/i386/kernel.c:356) Ke386L1CacheSize: 128kB (ke/i386/kernel.c:360) Ke386L2CacheSize: 256kB (io/pnpmgr.c:962) InstancePath is Root\PCI\0000 (io/pnpmgr.c:962) InstancePath is Root\Ne2000\0000 (io/pnpmgr.c:962) InstancePath is Root\LEGACY_UNKNOWN\0000 (io/pnpmgr.c:962) InstancePath is Root\LEGACY_UNKNOWN\0000 DriverBase for \SystemRoot\system32\drivers\pci.sys: dcc60000 Peripheral Component Interconnect Bus Driver (io/pnpmgr.c:962) InstancePath is PCI\VEN_1274&DEV_1371&SUBSYS_13711274&REV_02\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_1022&DEV_2000&SUBSYS_20001022&REV_10\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_104B&DEV_1040&SUBSYS_1040104B&REV_01\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_15AD&DEV_0405&SUBSYS_040515AD&REV_00\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_8086&DEV_7113&SUBSYS_197615AD&REV_08\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_8086&DEV_7111&SUBSYS_197615AD&REV_01\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_8086&DEV_7110&SUBSYS_197615AD&REV_08\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_8086&DEV_7191&SUBSYS_00000000&REV_01\0000 (io/pnpmgr.c:962) InstancePath is PCI\VEN_8086&DEV_7190&SUBSYS_197615AD&REV_01\0000 DriverBase for scsiport.sys: dcc69000 DriverBase for atapi.sys: dcc74000 (ldr/loader.c:1476) LdrPEGetExportByName(): failed to find NTOSKRNL.RtlConvertUlongToLargeInteger (ldr/loader.c:1561) Failed to import NTOSKRNL.RtlConvertUlongToLargeInteger from scsiport.sys KeBugCheckWithTf at ke/catch.c:164 Bug detected (code 1e param 0 0 0 0) KMODE_EXCEPTION_NOT_HANDLED Page Fault Exception: 14(0) Processor: 0 CS:EIP 8:58 cr2 58 cr3 2a7000 Proc: c0401010 Pid: 1 <SYSTEM> Thrd: c0406050 Tid: 1 DS 10 ES 10 FS 30 GS 10 EAX: 00000058 EBX: ccbd0998 ECX: 00000000 EDX: 00000000 EBP: c00b1c00 ESI: ccbd0998 ESP: c00b1ad0 EDI: 00000f00 EFLAGS: 00210286 kESP c00b1ad0 kernel stack base c00af000 Frames: <C007FE69> __________________________________ Do you Yahoo!? Check out the new Yahoo! Front Page. www.yahoo.com

20 years, 6 months

2
2
0 0

PATCH - Add support for Wine Regression Tests

by Steven Edwards

Hello, I am sending this in for review as I dont know how the CI system works and I dont know how to tie the Wine tests in to that system. I have already applyed Filips patch to wine/test.h that allows the test failure output to be redirected to DbgPrint so the CI system can log it. This patch only adds support for building the Wine tests in our build system. I have attached a sample set of tests (I dont know if they pass or not). just apply the winetest.patch and then dump the extract the tests folder in your reactos/lib/shlwapi folder. Now the next time you build shlwapi it will also build the tests. Thanks Steven __________________________________ Do you Yahoo!? Check out the new Yahoo! Front Page. www.yahoo.com

20 years, 6 months

1
0
0 0

RE: [ros-dev] KMODE_EXCEPTION after pressing CancelbuttoninTaskbarSettings dia

by Theodor Willax

Hi Ge, I've set my options accordingly now. Theo >From: "Ge van Geldorp" <gvg(a)reactos.com> >Reply-To: ReactOS Development List <ros-dev(a)reactos.com> >To: "'ReactOS Development List'" <ros-dev(a)reactos.com> >Subject: RE: [ros-dev] KMODE_EXCEPTION after pressing >CancelbuttoninTaskbarSettings dia >Date: Sat, 6 Nov 2004 23:49:21 +0100 > > > From: Theodor Willax > > > > Who will apply this patch to cvs > >I just did, thanks for debugging and providing the patch > > > and has it the right format? > >Not critical for a small patch like this, but it would be better if you >could use "cvs diff -u" to get some context next time. > >Gé van Geldorp. > > >_______________________________________________ >Ros-dev mailing list >Ros-dev(a)reactos.com >http://reactos.com:8080/mailman/listinfo/ros-dev _________________________________________________________________ Wenns mal schnell gehen soll. MSN Suche. http://search.msn.de/ Jetzt kostenlos nutzen und keine Zeit mehr verlieren.

20 years, 6 months

1
0
0 0

RE: [ros-dev] KMODE_EXCEPTION after pressing Cancel buttoninTaskbarSettings dia

by Theodor Willax

>--- Theodor Willax <theodor_willax(a)hotmail.com> wrote: > > thank's for the hint Ge. The patch below should fix this issue. Who > > will > > apply this patch to cvs and has it the right format? > >It needs to be in diff -u format. > >Thanks >Steven Here we are again with diff -u format cvs -z9 diff -u desktop.c (in directory D:\home\willaxt\Entwicklung\ros\reactos\subsys\win32k\ntuser\) Index: desktop.c =================================================================== RCS file: /CVS/ReactOS/reactos/subsys/win32k/ntuser/desktop.c,v retrieving revision 1.22 diff -u -r1.22 desktop.c --- desktop.c 28 Sep 2004 15:02:30 -0000 1.22 +++ desktop.c 6 Nov 2004 22:40:12 -0000 @@ -700,7 +700,7 @@ DPRINT("PaintDesktopVersionCallback ValueType=%d ValueLength=%d\n", ValueType, ValueLength); if (ValueType==REG_DWORD && ValueLength==sizeof(DWORD)) - *((DWORD*)Context) = *(DWORD*)ValueData; + *((DWORD*)EntryContext) = *(DWORD*)ValueData; return STATUS_SUCCESS; } _________________________________________________________________ Geschlossene Gesellschaft. MSN Hotmail mit McAfee® Anti-Virus. http://www.msn.de/email/antivirus/ Jetzt kostenlos anmelden und Viren bleiben vor der Tür!

20 years, 6 months

1
0
0 0

RE: [ros-dev] KMODE_EXCEPTION after pressing Cancel button inTaskbarSettings dia

by Theodor Willax

Hi, > > How could I debug this error? > >You can use the map files to find out where things went wrong. On the BSOD >you attached you can find that the kernel crashed at <win32k.sys: 142f8>. >So >you can go to subsys/win32k/win32k.map and find (relative) address 142f8. >If >you compiled with DBG := 1 in your config file this should give you the >exact source line where the crash occurred. > >If you need a stack backtrace, use the values on the "Frames:" line in the >same way. > >Gé van Geldorp. thank's for the hint Ge. The patch below should fix this issue. Who will apply this patch to cvs and has it the right format? Theo cvs -z9 diff desktop.c (in directory D:\home\willaxt\Entwicklung\ros\reactos\subsys\win32k\ntuser\) Index: desktop.c =================================================================== RCS file: /CVS/ReactOS/reactos/subsys/win32k/ntuser/desktop.c,v retrieving revision 1.22 diff -r1.22 desktop.c 703c703 < *((DWORD*)Context) = *(DWORD*)ValueData; --- > *((DWORD*)EntryContext) = *(DWORD*)ValueData; _________________________________________________________________ Die rote Karte für lästige E-Mails. MSN Hotmail mit Junk-Mail-Filter. http://www.msn.de/antispam/prevention/junkmailfilter Jetzt kostenlos anmelden!

20 years, 6 months

3
2
0 0

Test stub generator

by art yerkes

I wrote a small tool to generate the tests/stubs.tst file for a component. http://mok.lvcm.com/cgi-bin/reactos/roswiki?TestStubsGenerator -- Here's a simple experiment. Stand on a train track between two locomotives which are pushing on you with equal force in opposite directions. You will exhibit no net motion. None the less, you may soon begin to notice that something important is happening. -- Robert Stirniman

20 years, 6 months

1
0
0 0

Re: Re: [ros-dev] ReactOS setup CD-ROM crashes on old hardware.

by christoph＠apiviewer.de

> Thanks for reporting. Such error was already reported by one person on > our IRC channel, but AFAIK wasn't solved yet. That was me :-) Please notice me if a newer freeloader is available for testing... Christoph von Wittich

20 years, 6 months

1
0
0 0

ROS Build Failure with Explorer

by Michael B. Trausch

Okay, I'm just trying to build the system to see what's going on with it, and thinking about doing some application related work, maybe, at some point, but in any case, I can't get ReactOS to build. The build dies when it reaches explorer.cpp, and you can see in the attached error log (explorer-build.log-2). The version of gcc used was: fd0man@fd0man:~> i386-mingw32msvc-gcc --version i386-mingw32msvc-gcc (GCC) 3.4.1 (mingw special) Copyright (C) 2004 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. The compiler is a Linux-to-Win32 cross compiler. The CVS tree is fully up-to-date as of right this minute, as I just ran the update before I sent this off and didn't see any changes occur. I don't have the ability to poke around with it at the moment, nor do I have the slightest clue of what's causing it... it seems that it's saying something about an overloaded function - when it's a completely different function that's called on the line it's complaining about, anyway, just doesn't seem to make sense to me. *shrugs* Later, Michael

20 years, 6 months

2
1
0 0

Why Ekush can't GPL

by Alex Ionescu

Hi, I'm sure you've all heard about ekush... I recently read this and though it was a pearl: "Licensing is one of the problems the *Ekush* team is expecting to face. As the project is not based in the US, *Ekush* OS will not be able to obtain the license banner of General Public License (GPL), the US-based licensing company." It just made me burst out laughing :) Best regards, Alex Ionescu

20 years, 6 months

1
0
0 0

The ne2000 driver

by art yerkes

I'm continuing to have trouble with the ne2000 driver, and I'm honestly not sure how to fix it. I've been poking around a little, so I'll put a diff from my tree as well. Basically, the problem i see is that BusType and BusNumber from IoGetDeviceProperty(..., DevicePropertyLegacyBusType, ..., &Adapter->BusType, ...) and IoGetDeviceProperty(..., DevicePropertyLegacyBusNumber, ...) are always -1 in drivers/net/ndis/ndis/miniport.c, line 1404. Even when i force these values to sensible ones (Isa, 0), the ne2000 gives 0xff for all the bytes of the mac address. I should mention that I'm using qemu. I've tried my trusty 32-bit 0.5.5 build, 32-bit 0.6.0, and 64-bit 0.6.0 builds. http://64.81.145.152/~arty/ne2000-goodies.zip --> arty/arty-poke.diff --> arty/run-log.txt --> arty/qemu-0.* (various versions of qemu i tested if it helps) -- Here's a simple experiment. Stand on a train track between two locomotives which are pushing on you with equal force in opposite directions. You will exhibit no net motion. None the less, you may soon begin to notice that something important is happening. -- Robert Stirniman

20 years, 6 months

2
2
0 0

ros-cvs list question

by Hartmut Birr

Hi, it is possible, to add the version of a commited file? - Hartmut

20 years, 6 months

1
0
0 0

Do you need this file?

by Jakob Eriksson

http://www.acc.umu.se/~bosse/ntifs.h

20 years, 6 months

2
1
0 0

KMODE_EXCEPTION after pressing Cancel button in Taskbar Settings dialog

by Theodor Willax

Hello, I'm getting repeatedly a KMODE_EXCEPTION when pressing the Cancel button in the Taskbar Settings dialog. See both screenshots attached. I got this blue screen with several versions of the 0.2.5 cvs tree on my virtual pc. The last build I did with the cvs tree from yesterday evening. How could I debug this error? Thanks, Theo _________________________________________________________________ Machen Sie lästigen E-Mails ein Ende. MSN Hotmail mit Junk-Mail-Filter. http://www.msn.de/antispam/prevention/junkmailfilter Jetzt kostenlos anmelden!

20 years, 6 months

2
1
0 0

Re: [ros-dev] connect gdb to named pipe in ms virtual pc session

by Theodor Willax

Hello Filip, I see what your patch should achieve. But pipetunnel doesn't work although I defined out the serial port detection code according to your patch. I will post if I'm successful in getting pipetunnel to work on virtual pc. Thanks, Theo >From: Filip Navara <xnavara(a)volny.cz> >Reply-To: ReactOS Development List <ros-dev(a)reactos.com> >To: ReactOS Development List <ros-dev(a)reactos.com> >Subject: Re: [ros-dev] connect gdb to named pipe in ms virtual pc session >Date: Wed, 03 Nov 2004 19:37:21 +0100 > >Theodor Willax wrote: > >>a) Do I need to compile the cvs snapshot with special make options to >>debug trough serial ports? > >Generally no, but for Virtual PC you need to get do a CVS build of HAL with >hacked serial port detection. The standard detection approach doesn't work >in VPC because of "bug" in the emulated hardware. (Yes, it's really broken. >It works in Linux because it assumes that two ports always exist and skips >the detection. See the Linux source for details.) I don't have the original >patch I used while debugging the S3 driver under VPC anymore, but the >attached one should work (I hope)... > >BTW, I'm not able to answer your other questions, sorry. > >Regards, >Filip > >Index: hal/halx86/kdbg.c >=================================================================== >RCS file: /CVS/ReactOS/reactos/hal/halx86/kdbg.c,v >retrieving revision 1.8 >diff -u -r1.8 kdbg.c >--- hal/halx86/kdbg.c 29 Apr 2004 17:06:21 -0000 1.8 >+++ hal/halx86/kdbg.c 3 Nov 2004 18:32:30 -0000 >@@ -102,6 +102,7 @@ > /* clear all modem output bits */ > WRITE_PORT_UCHAR (SER_MCR(BaseAddress), 0x10); > >+#if 0 > /* read the Modem Status Register */ > msr = READ_PORT_UCHAR (SER_MSR(BaseAddress)); > >@@ -123,6 +124,7 @@ > */ > if ((msr & 0xF0) == 0xF0) > { >+#endif > /* > * setup a resonable state for the port: > * enable fifo and clear recieve/transmit buffers >@@ -133,8 +135,10 @@ > READ_PORT_UCHAR (SER_RBR(BaseAddress)); > WRITE_PORT_UCHAR (SER_IER(BaseAddress), 0); > found = TRUE; >+#if 0 > } > } >+#endif > > /* restore MCR */ > WRITE_PORT_UCHAR (SER_MCR(BaseAddress), mcr); >_______________________________________________ >Ros-dev mailing list >Ros-dev(a)reactos.com >http://reactos.com:8080/mailman/listinfo/ros-dev _________________________________________________________________ Ungestört surfen. MSN Toolbar mit Pop-up-Blocker. http://toolbar.msn.de/ Jetzt kostenlos downloaden!

20 years, 6 months

1
0
0 0

Using zones in NTOS

by Aliberti Emanuele

What are the advantages in using zones when allocating memory in kernel mode? Are there other ways to pre-allocate an address range to be used for small memory objects management (NP pool)? Emanuele

20 years, 6 months

3
6
0 0

RE: [ros-dev] connect gdb to named pipe in ms virtual pc session

by Theodor Willax

>From: "Theodor Willax" <theodor_willax(a)hotmail.com> >Reply-To: ReactOS Development List <ros-dev(a)reactos.com> >To: ros-dev(a)reactos.com >Subject: [ros-dev] connect gdb to named pipe in ms virtual pc session >Date: Wed, 03 Nov 2004 11:40:35 +0000 > >Hi, > >could anybody tell me how to connect gdb.exe to my named pipe >"\\.\pipe\roscom2" of a virtual pc session (on virtual com2 port). I'm >running gdb on Win XP SP2 and use ros 0.2.5-cvs last updated 11/03/04. >I set up "pipetunnel.exe roscom2" in one command prompt window, started >virtual pc session in debug mode and now try to connect gdb to pipetunnel. >pipetunnel is now listening on tcp port 9999. > >Thanks, >Theo > Hi, it's me again. Some more questions: a) Do I need to compile the cvs snapshot with special make options to debug trough serial ports? b) Works the named pipe approach with ms virtual pc? c) Do I need a special gdb version? I've this one: D:\home\willaxt\Entwicklung\ros>gdb -version GNU gdb 5.2.1 Copyright 2002 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "i686-pc-mingw32". Many thank's, Theo _________________________________________________________________ Geschlossene Gesellschaft. MSN Hotmail mit McAfee® Anti-Virus. http://www.msn.de/email/antivirus/ Jetzt kostenlos anmelden und Viren bleiben vor der Tür!

20 years, 6 months

2
1
0 0

address of a variable

by Hartmut Birr

Hi, if I have address of a variable in the bss section from ntoskrnl, how can I find out the name? Searching into the map file doesn't help. Possible the address is a pointer into a structure. - Hartmut

20 years, 6 months

2
4
0 0

connect gdb to named pipe in ms virtual pc session

by Theodor Willax

Hi, could anybody tell me how to connect gdb.exe to my named pipe "\\.\pipe\roscom2" of a virtual pc session (on virtual com2 port). I'm running gdb on Win XP SP2 and use ros 0.2.5-cvs last updated 11/03/04. I set up "pipetunnel.exe roscom2" in one command prompt window, started virtual pc session in debug mode and now try to connect gdb to pipetunnel. pipetunnel is now listening on tcp port 9999. Thanks, Theo _________________________________________________________________ Geschlossene Gesellschaft. MSN Hotmail mit McAfee® Anti-Virus. http://www.msn.de/email/antivirus/ Jetzt kostenlos anmelden und Viren bleiben vor der Tür!

20 years, 6 months

1
0
0 0

Re: [ros-cvs] CVS Update: reactos

by Alex Ionescu

weiden(a)cvs.reactos.com wrote: > 1. reformatted the code so human beings can read it > > Best CVS log comment. Ever.

20 years, 6 months

2
1
0 0

Re: [ros-cvs] CVS Update: reactos

by Filip Navara

Hartmut Birr wrote: >CVSROOT: /CVS/ReactOS >Module name: reactos >Repository: reactos/ntoskrnl/ke/ >Changes by: hbirr(a)mok.osexperts.com 04/10/31 07:24:08 > >Modified files: > reactos/ntoskrnl/ke/: timer.c > >Log message: > - The quantum of the idle thread must also end. > > I thought that the NtYieldExecution call in PsIdleThreadMain should be sufficient, isn't it? Regards, Filip

20 years, 6 months

2
1
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Ros-dev