You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
208 lines
4.8 KiB
208 lines
4.8 KiB
/*++
|
|
|
|
Copyright (c) 1990-2001 Microsoft Corporation
|
|
|
|
Module Name:
|
|
|
|
kdmove.c
|
|
|
|
Abstract:
|
|
|
|
This module contains code to implement the portable kernel debugger
|
|
memory mover.
|
|
|
|
Author:
|
|
|
|
Mark Lucovsky (markl) 31-Aug-1990
|
|
|
|
Revision History:
|
|
|
|
--*/
|
|
|
|
#include "kdp.h"
|
|
|
|
#ifdef ALLOC_PRAGMA
|
|
#pragma alloc_text(PAGEKD, KdpQuickMoveMemory)
|
|
#pragma alloc_text(PAGEKD, KdpCopyMemoryChunks)
|
|
#endif
|
|
|
|
VOID
|
|
KdpQuickMoveMemory (
|
|
IN PCHAR Destination,
|
|
IN PCHAR Source,
|
|
IN ULONG Length
|
|
)
|
|
|
|
/*++
|
|
|
|
Routine Description:
|
|
|
|
This routine does the exact same thing as RtlCopyMemory, BUT it is
|
|
private to the debugger. This allows folks to set breakpoints and
|
|
watch points in RtlCopyMemory without risk of recursive debugger
|
|
entry and the accompanying hang.
|
|
|
|
N.B. UNLIKE KdpCopyMemoryChunks, this routine does NOT check for
|
|
accessibility and may fault! Use it ONLY in the debugger and ONLY
|
|
where you could use RtlCopyMemory.
|
|
|
|
Arguments:
|
|
|
|
Destination - Supplies a pointer to destination of the move operation.
|
|
|
|
Source - Supplies a pointer to the source of the move operation.
|
|
|
|
Length - Supplies the length of the move operation.
|
|
|
|
Return Value:
|
|
|
|
None.
|
|
|
|
--*/
|
|
{
|
|
while (Length > 0) {
|
|
*Destination = *Source;
|
|
Destination++;
|
|
Source++;
|
|
Length--;
|
|
}
|
|
}
|
|
|
|
NTSTATUS
|
|
KdpCopyMemoryChunks(
|
|
ULONG64 Address,
|
|
PVOID Buffer,
|
|
ULONG TotalSize,
|
|
ULONG ChunkSize,
|
|
ULONG Flags,
|
|
PULONG ActualSize OPTIONAL
|
|
)
|
|
|
|
/*++
|
|
|
|
Routine Description:
|
|
|
|
Copies memory to/from a buffer to/from a system address.
|
|
The address can be physical or virtual.
|
|
|
|
The buffer is assumed to be valid for the duration of this call.
|
|
|
|
Arguments:
|
|
|
|
Address - System address.
|
|
|
|
Buffer - Buffer to read from or write to.
|
|
|
|
TotalSize - Number of bytes to read/write.
|
|
|
|
ChunkSize - Maximum single item transfer size, must
|
|
be 1, 2, 4 or 8.
|
|
0 means choose a default.
|
|
|
|
Flags - MMDBG_COPY flags for MmDbgCopyMemory.
|
|
|
|
ActualSize - Number of bytes actually read/written.
|
|
|
|
Return Value:
|
|
|
|
NTSTATUS
|
|
|
|
--*/
|
|
|
|
{
|
|
ULONG Length;
|
|
ULONG CopyChunk;
|
|
NTSTATUS Status;
|
|
#if defined(_IA64_)
|
|
ULONG64 AddressStart = Address;
|
|
#endif
|
|
|
|
if (ChunkSize > MMDBG_COPY_MAX_SIZE) {
|
|
ChunkSize = MMDBG_COPY_MAX_SIZE;
|
|
} else if (ChunkSize == 0) {
|
|
// Default to 4 byte chunks as that's
|
|
// what the previous code did.
|
|
ChunkSize = 4;
|
|
}
|
|
|
|
//
|
|
// MmDbgCopyMemory only copies a single aligned chunk at a
|
|
// time. It is Kd's responsibility to chunk up a larger
|
|
// request for individual copy requests. This gives Kd
|
|
// the flexibility to pick a chunk size and also frees
|
|
// Mm from having to worry about more than a page at a time.
|
|
// Additionally, it is important that we access memory with the
|
|
// largest size possible because we could be accessing
|
|
// memory-mapped I/O space.
|
|
//
|
|
|
|
Length = TotalSize;
|
|
CopyChunk = 1;
|
|
|
|
while (Length > 0) {
|
|
|
|
// Expand the chunk size as long as:
|
|
// We haven't hit the chunk limit.
|
|
// We have enough data left.
|
|
// The address is properly aligned.
|
|
while (CopyChunk < ChunkSize &&
|
|
(CopyChunk << 1) <= Length &&
|
|
(Address & ((CopyChunk << 1) - 1)) == 0) {
|
|
CopyChunk <<= 1;
|
|
}
|
|
|
|
// Shrink the chunk size to fit the available data.
|
|
while (CopyChunk > Length) {
|
|
CopyChunk >>= 1;
|
|
}
|
|
|
|
Status = MmDbgCopyMemory(Address, Buffer, CopyChunk, Flags);
|
|
|
|
if (!NT_SUCCESS(Status)) {
|
|
break;
|
|
}
|
|
|
|
Address += CopyChunk;
|
|
Buffer = (PVOID)((PUCHAR)Buffer + CopyChunk);
|
|
Length -= CopyChunk;
|
|
}
|
|
|
|
if (ActualSize)
|
|
{
|
|
*ActualSize = TotalSize - Length;
|
|
}
|
|
|
|
//
|
|
// Flush the instruction cache in case the write was into the instruction
|
|
// stream. Only do this when writing into the kernel address space,
|
|
// and if any bytes were actually written
|
|
//
|
|
|
|
if ((Flags & MMDBG_COPY_WRITE) &&
|
|
Length < TotalSize) {
|
|
|
|
#if defined(_IA64_)
|
|
|
|
//
|
|
// KeSweepCurrentIcacheRange requires a valid virtual address.
|
|
// It is used because KeSweepCurrentICache does not work until
|
|
// the HAL as been initialized.
|
|
//
|
|
|
|
if (Flags & MMDBG_COPY_PHYSICAL) {
|
|
|
|
KeSweepCurrentIcache();
|
|
|
|
} else{
|
|
|
|
KeSweepCurrentIcacheRange((PVOID)AddressStart, TotalSize - Length);
|
|
}
|
|
|
|
#else
|
|
KeSweepCurrentIcache();
|
|
#endif
|
|
|
|
}
|
|
|
|
return Length != 0 ? STATUS_UNSUCCESSFUL : STATUS_SUCCESS;
|
|
}
|