QED.cpp@ 69500

Last change on this file since 69500 was 69500, checked in by vboxsync, 7 years ago
*: scm --update-copyright-year
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 86.4 KB

Line
1	/* $Id: QED.cpp 69500 2017-10-28 15:14:05Z vboxsync $ */
2	/** @file
3	* QED - QED Disk image.
4	*/
5
6	/*
7	* Copyright (C) 2011-2017 Oracle Corporation
8	*
9	* This file is part of VirtualBox Open Source Edition (OSE), as
10	* available from http://www.virtualbox.org. This file is free software;
11	* you can redistribute it and/or modify it under the terms of the GNU
12	* General Public License (GPL) as published by the Free Software
13	* Foundation, in version 2 as it comes in the "COPYING" file of the
14	* VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15	* hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16	*/
17
18
19	/*********************************************************************************************************************************
20	* Header Files *
21	*********************************************************************************************************************************/
22	#define LOG_GROUP LOG_GROUP_VD_QED
23	#include <VBox/vd-plugin.h>
24	#include <VBox/err.h>
25
26	#include <VBox/log.h>
27	#include <iprt/asm.h>
28	#include <iprt/assert.h>
29	#include <iprt/string.h>
30	#include <iprt/alloc.h>
31	#include <iprt/path.h>
32	#include <iprt/list.h>
33
34	#include "VDBackends.h"
35	#include "VDBackendsInline.h"
36
37	/**
38	* The QED backend implements support for the qemu enhanced disk format (short QED)
39	* The specification for the format is available under http://wiki.qemu.org/Features/QED/Specification
40	*
41	* Missing things to implement:
42	* - compaction
43	* - resizing which requires block relocation (very rare case)
44	*/
45
46
47	/*********************************************************************************************************************************
48	* Structures in a QED image, little endian *
49	*********************************************************************************************************************************/
50
51	#pragma pack(1)
52	typedef struct QedHeader
53	{
54	/** Magic value. */
55	uint32_t u32Magic;
56	/** Cluster size in bytes. */
57	uint32_t u32ClusterSize;
58	/** Size of L1 and L2 tables in clusters. */
59	uint32_t u32TableSize;
60	/** size of this header structure in clusters. */
61	uint32_t u32HeaderSize;
62	/** Features used for the image. */
63	uint64_t u64FeatureFlags;
64	/** Compatibility features used for the image. */
65	uint64_t u64CompatFeatureFlags;
66	/** Self resetting feature bits. */
67	uint64_t u64AutoresetFeatureFlags;
68	/** Offset of the L1 table in bytes. */
69	uint64_t u64OffL1Table;
70	/** Logical image size as seen by the guest. */
71	uint64_t u64Size;
72	/** Offset of the backing filename in bytes. */
73	uint32_t u32OffBackingFilename;
74	/** Size of the backing filename. */
75	uint32_t u32BackingFilenameSize;
76	} QedHeader;
77	#pragma pack()
78	/** Pointer to a on disk QED header. */
79	typedef QedHeader *PQedHeader;
80
81	/** QED magic value. */
82	#define QED_MAGIC UINT32_C(0x00444551) /* QED\0 */
83	/** Cluster size minimum. */
84	#define QED_CLUSTER_SIZE_MIN RT_BIT(12)
85	/** Cluster size maximum. */
86	#define QED_CLUSTER_SIZE_MAX RT_BIT(26)
87	/** L1 and L2 Table size minimum. */
88	#define QED_TABLE_SIZE_MIN 1
89	/** L1 and L2 Table size maximum. */
90	#define QED_TABLE_SIZE_MAX 16
91
92	/** QED default cluster size when creating an image. */
93	#define QED_CLUSTER_SIZE_DEFAULT (64 * _1K)
94	/** The default table size in clusters. */
95	#define QED_TABLE_SIZE_DEFAULT 4
96
97	/** Feature flags.
98	* @{
99	*/
100	/** Image uses a backing file to provide data for unallocated clusters. */
101	#define QED_FEATURE_BACKING_FILE RT_BIT_64(0)
102	/** Image needs checking before use. */
103	#define QED_FEATURE_NEED_CHECK RT_BIT_64(1)
104	/** Don't probe for format of the backing file, treat as raw image. */
105	#define QED_FEATURE_BACKING_FILE_NO_PROBE RT_BIT_64(2)
106	/** Mask of valid features. */
107	#define QED_FEATURE_MASK (QED_FEATURE_BACKING_FILE \| QED_FEATURE_NEED_CHECK \| QED_FEATURE_BACKING_FILE_NO_PROBE)
108	/** @} */
109
110	/** Compatibility feature flags.
111	* @{
112	*/
113	/** Mask of valid compatibility features. */
114	#define QED_COMPAT_FEATURE_MASK (0)
115	/** @} */
116
117	/** Autoreset feature flags.
118	* @{
119	*/
120	/** Mask of valid autoreset features. */
121	#define QED_AUTORESET_FEATURE_MASK (0)
122	/** @} */
123
124
125	/*********************************************************************************************************************************
126	* Constants And Macros, Structures and Typedefs *
127	*********************************************************************************************************************************/
128
129	/**
130	* QED L2 cache entry.
131	*/
132	typedef struct QEDL2CACHEENTRY
133	{
134	/** List node for the search list. */
135	RTLISTNODE NodeSearch;
136	/** List node for the LRU list. */
137	RTLISTNODE NodeLru;
138	/** Reference counter. */
139	uint32_t cRefs;
140	/** The offset of the L2 table, used as search key. */
141	uint64_t offL2Tbl;
142	/** Pointer to the cached L2 table. */
143	uint64_t *paL2Tbl;
144	} QEDL2CACHEENTRY, *PQEDL2CACHEENTRY;
145
146	/** Maximum amount of memory the cache is allowed to use. */
147	#define QED_L2_CACHE_MEMORY_MAX (2*_1M)
148
149	/**
150	* QED image data structure.
151	*/
152	typedef struct QEDIMAGE
153	{
154	/** Image name. */
155	const char *pszFilename;
156	/** Storage handle. */
157	PVDIOSTORAGE pStorage;
158
159	/** Pointer to the per-disk VD interface list. */
160	PVDINTERFACE pVDIfsDisk;
161	/** Pointer to the per-image VD interface list. */
162	PVDINTERFACE pVDIfsImage;
163	/** Error interface. */
164	PVDINTERFACEERROR pIfError;
165	/** I/O interface. */
166	PVDINTERFACEIOINT pIfIo;
167
168	/** Open flags passed by VBoxHD layer. */
169	unsigned uOpenFlags;
170	/** Image flags defined during creation or determined during open. */
171	unsigned uImageFlags;
172	/** Total size of the image. */
173	uint64_t cbSize;
174	/** Physical geometry of this image. */
175	VDGEOMETRY PCHSGeometry;
176	/** Logical geometry of this image. */
177	VDGEOMETRY LCHSGeometry;
178
179	/** Filename of the backing file if any. */
180	char *pszBackingFilename;
181	/** Offset of the filename in the image. */
182	uint32_t offBackingFilename;
183	/** Size of the backing filename excluding \0. */
184	uint32_t cbBackingFilename;
185
186	/** Size of the image, multiple of clusters. */
187	uint64_t cbImage;
188	/** Cluster size in bytes. */
189	uint32_t cbCluster;
190	/** Number of entries in the L1 and L2 table. */
191	uint32_t cTableEntries;
192	/** Size of an L1 or L2 table rounded to the next cluster size. */
193	uint32_t cbTable;
194	/** Pointer to the L1 table. */
195	uint64_t *paL1Table;
196	/** Offset of the L1 table. */
197	uint64_t offL1Table;
198
199	/** Offset mask for a cluster. */
200	uint64_t fOffsetMask;
201	/** L1 table mask to get the L1 index. */
202	uint64_t fL1Mask;
203	/** Number of bits to shift to get the L1 index. */
204	uint32_t cL1Shift;
205	/** L2 table mask to get the L2 index. */
206	uint64_t fL2Mask;
207	/** Number of bits to shift to get the L2 index. */
208	uint32_t cL2Shift;
209
210	/** Pointer to the L2 table we are currently allocating
211	* (can be only one at a time). */
212	PQEDL2CACHEENTRY pL2TblAlloc;
213
214	/** Memory occupied by the L2 table cache. */
215	size_t cbL2Cache;
216	/** The sorted L2 entry list used for searching. */
217	RTLISTNODE ListSearch;
218	/** The LRU L2 entry list used for eviction. */
219	RTLISTNODE ListLru;
220	/** The static region list. */
221	VDREGIONLIST RegionList;
222	} QEDIMAGE, *PQEDIMAGE;
223
224	/**
225	* State of the async cluster allocation.
226	*/
227	typedef enum QEDCLUSTERASYNCALLOCSTATE
228	{
229	/** Invalid. */
230	QEDCLUSTERASYNCALLOCSTATE_INVALID = 0,
231	/** L2 table allocation. */
232	QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC,
233	/** Link L2 table into L1. */
234	QEDCLUSTERASYNCALLOCSTATE_L2_LINK,
235	/** Allocate user data cluster. */
236	QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC,
237	/** Link user data cluster. */
238	QEDCLUSTERASYNCALLOCSTATE_USER_LINK,
239	/** 32bit blowup. */
240	QEDCLUSTERASYNCALLOCSTATE_32BIT_HACK = 0x7fffffff
241	} QEDCLUSTERASYNCALLOCSTATE, *PQEDCLUSTERASYNCALLOCSTATE;
242
243	/**
244	* Data needed to track async cluster allocation.
245	*/
246	typedef struct QEDCLUSTERASYNCALLOC
247	{
248	/** The state of the cluster allocation. */
249	QEDCLUSTERASYNCALLOCSTATE enmAllocState;
250	/** Old image size to rollback in case of an error. */
251	uint64_t cbImageOld;
252	/** L1 index to link if any. */
253	uint32_t idxL1;
254	/** L2 index to link, required in any case. */
255	uint32_t idxL2;
256	/** Start offset of the allocated cluster. */
257	uint64_t offClusterNew;
258	/** L2 cache entry if a L2 table is allocated. */
259	PQEDL2CACHEENTRY pL2Entry;
260	/** Number of bytes to write. */
261	size_t cbToWrite;
262	} QEDCLUSTERASYNCALLOC, *PQEDCLUSTERASYNCALLOC;
263
264
265	/*********************************************************************************************************************************
266	* Static Variables *
267	*********************************************************************************************************************************/
268
269	/** NULL-terminated array of supported file extensions. */
270	static const VDFILEEXTENSION s_aQedFileExtensions[] =
271	{
272	{"qed", VDTYPE_HDD},
273	{NULL, VDTYPE_INVALID}
274	};
275
276
277	/*********************************************************************************************************************************
278	* Internal Functions *
279	*********************************************************************************************************************************/
280
281	/**
282	* Converts the image header to the host endianess and performs basic checks.
283	*
284	* @returns Whether the given header is valid or not.
285	* @param pHeader Pointer to the header to convert.
286	*/
287	static bool qedHdrConvertToHostEndianess(PQedHeader pHeader)
288	{
289	pHeader->u32Magic = RT_LE2H_U32(pHeader->u32Magic);
290	pHeader->u32ClusterSize = RT_LE2H_U32(pHeader->u32ClusterSize);
291	pHeader->u32TableSize = RT_LE2H_U32(pHeader->u32TableSize);
292	pHeader->u32HeaderSize = RT_LE2H_U32(pHeader->u32HeaderSize);
293	pHeader->u64FeatureFlags = RT_LE2H_U64(pHeader->u64FeatureFlags);
294	pHeader->u64CompatFeatureFlags = RT_LE2H_U64(pHeader->u64CompatFeatureFlags);
295	pHeader->u64AutoresetFeatureFlags = RT_LE2H_U64(pHeader->u64AutoresetFeatureFlags);
296	pHeader->u64OffL1Table = RT_LE2H_U64(pHeader->u64OffL1Table);
297	pHeader->u64Size = RT_LE2H_U64(pHeader->u64Size);
298	pHeader->u32OffBackingFilename = RT_LE2H_U32(pHeader->u32OffBackingFilename);
299	pHeader->u32BackingFilenameSize = RT_LE2H_U32(pHeader->u32BackingFilenameSize);
300
301	if (RT_UNLIKELY(pHeader->u32Magic != QED_MAGIC))
302	return false;
303	if (RT_UNLIKELY( pHeader->u32ClusterSize < QED_CLUSTER_SIZE_MIN
304	\|\| pHeader->u32ClusterSize > QED_CLUSTER_SIZE_MAX))
305	return false;
306	if (RT_UNLIKELY( pHeader->u32TableSize < QED_TABLE_SIZE_MIN
307	\|\| pHeader->u32TableSize > QED_TABLE_SIZE_MAX))
308	return false;
309	if (RT_UNLIKELY(pHeader->u64Size % 512 != 0))
310	return false;
311	if (RT_UNLIKELY( pHeader->u64FeatureFlags & QED_FEATURE_BACKING_FILE
312	&& ( pHeader->u32BackingFilenameSize == 0
313	\|\| pHeader->u32BackingFilenameSize == UINT32_MAX)))
314	return false;
315
316	return true;
317	}
318
319	/**
320	* Creates a QED header from the given image state.
321	*
322	* @returns nothing.
323	* @param pImage Image instance data.
324	* @param pHeader Pointer to the header to convert.
325	*/
326	static void qedHdrConvertFromHostEndianess(PQEDIMAGE pImage, PQedHeader pHeader)
327	{
328	pHeader->u32Magic = RT_H2LE_U32(QED_MAGIC);
329	pHeader->u32ClusterSize = RT_H2LE_U32(pImage->cbCluster);
330	pHeader->u32TableSize = RT_H2LE_U32(pImage->cbTable / pImage->cbCluster);
331	pHeader->u32HeaderSize = RT_H2LE_U32(1);
332	pHeader->u64FeatureFlags = RT_H2LE_U64(pImage->pszBackingFilename ? QED_FEATURE_BACKING_FILE : UINT64_C(0));
333	pHeader->u64CompatFeatureFlags = RT_H2LE_U64(UINT64_C(0));
334	pHeader->u64AutoresetFeatureFlags = RT_H2LE_U64(UINT64_C(0));
335	pHeader->u64OffL1Table = RT_H2LE_U64(pImage->offL1Table);
336	pHeader->u64Size = RT_H2LE_U64(pImage->cbSize);
337	pHeader->u32OffBackingFilename = RT_H2LE_U32(pImage->offBackingFilename);
338	pHeader->u32BackingFilenameSize = RT_H2LE_U32(pImage->cbBackingFilename);
339	}
340
341	/**
342	* Convert table entries from little endian to host endianess.
343	*
344	* @returns nothing.
345	* @param paTbl Pointer to the table.
346	* @param cEntries Number of entries in the table.
347	*/
348	static void qedTableConvertToHostEndianess(uint64_t *paTbl, uint32_t cEntries)
349	{
350	while(cEntries-- > 0)
351	{
352	paTbl = RT_LE2H_U64(paTbl);
353	paTbl++;
354	}
355	}
356
357	#if defined(RT_BIG_ENDIAN)
358	/**
359	* Convert table entries from host to little endian format.
360	*
361	* @returns nothing.
362	* @param paTblImg Pointer to the table which will store the little endian table.
363	* @param paTbl The source table to convert.
364	* @param cEntries Number of entries in the table.
365	*/
366	static void qedTableConvertFromHostEndianess(uint64_t paTblImg, uint64_t paTbl,
367	uint32_t cEntries)
368	{
369	while(cEntries-- > 0)
370	{
371	paTblImg = RT_H2LE_U64(paTbl);
372	paTbl++;
373	paTblImg++;
374	}
375	}
376	#endif
377
378	/**
379	* Creates the L2 table cache.
380	*
381	* @returns VBox status code.
382	* @param pImage The image instance data.
383	*/
384	static int qedL2TblCacheCreate(PQEDIMAGE pImage)
385	{
386	pImage->cbL2Cache = 0;
387	RTListInit(&pImage->ListSearch);
388	RTListInit(&pImage->ListLru);
389
390	return VINF_SUCCESS;
391	}
392
393	/**
394	* Destroys the L2 table cache.
395	*
396	* @returns nothing.
397	* @param pImage The image instance data.
398	*/
399	static void qedL2TblCacheDestroy(PQEDIMAGE pImage)
400	{
401	PQEDL2CACHEENTRY pL2Entry;
402	PQEDL2CACHEENTRY pL2Next;
403	RTListForEachSafe(&pImage->ListSearch, pL2Entry, pL2Next, QEDL2CACHEENTRY, NodeSearch)
404	{
405	Assert(!pL2Entry->cRefs);
406
407	RTListNodeRemove(&pL2Entry->NodeSearch);
408	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
409	RTMemFree(pL2Entry);
410	}
411
412	pImage->cbL2Cache = 0;
413	RTListInit(&pImage->ListSearch);
414	RTListInit(&pImage->ListLru);
415	}
416
417	/**
418	* Returns the L2 table matching the given offset or NULL if none could be found.
419	*
420	* @returns Pointer to the L2 table cache entry or NULL.
421	* @param pImage The image instance data.
422	* @param offL2Tbl Offset of the L2 table to search for.
423	*/
424	static PQEDL2CACHEENTRY qedL2TblCacheRetain(PQEDIMAGE pImage, uint64_t offL2Tbl)
425	{
426	if ( pImage->pL2TblAlloc
427	&& pImage->pL2TblAlloc->offL2Tbl == offL2Tbl)
428	{
429	pImage->pL2TblAlloc->cRefs++;
430	return pImage->pL2TblAlloc;
431	}
432
433	PQEDL2CACHEENTRY pL2Entry;
434	RTListForEach(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch)
435	{
436	if (pL2Entry->offL2Tbl == offL2Tbl)
437	break;
438	}
439
440	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
441	{
442	/* Update LRU list. */
443	RTListNodeRemove(&pL2Entry->NodeLru);
444	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
445	pL2Entry->cRefs++;
446	return pL2Entry;
447	}
448	else
449	return NULL;
450	}
451
452	/**
453	* Releases a L2 table cache entry.
454	*
455	* @returns nothing.
456	* @param pL2Entry The L2 cache entry.
457	*/
458	static void qedL2TblCacheEntryRelease(PQEDL2CACHEENTRY pL2Entry)
459	{
460	Assert(pL2Entry->cRefs > 0);
461	pL2Entry->cRefs--;
462	}
463
464	/**
465	* Allocates a new L2 table from the cache evicting old entries if required.
466	*
467	* @returns Pointer to the L2 cache entry or NULL.
468	* @param pImage The image instance data.
469	*/
470	static PQEDL2CACHEENTRY qedL2TblCacheEntryAlloc(PQEDIMAGE pImage)
471	{
472	PQEDL2CACHEENTRY pL2Entry = NULL;
473
474	if (pImage->cbL2Cache + pImage->cbTable <= QED_L2_CACHE_MEMORY_MAX)
475	{
476	/* Add a new entry. */
477	pL2Entry = (PQEDL2CACHEENTRY)RTMemAllocZ(sizeof(QEDL2CACHEENTRY));
478	if (pL2Entry)
479	{
480	pL2Entry->paL2Tbl = (uint64_t *)RTMemPageAllocZ(pImage->cbTable);
481	if (RT_UNLIKELY(!pL2Entry->paL2Tbl))
482	{
483	RTMemFree(pL2Entry);
484	pL2Entry = NULL;
485	}
486	else
487	{
488	pL2Entry->cRefs = 1;
489	pImage->cbL2Cache += pImage->cbTable;
490	}
491	}
492	}
493	else
494	{
495	/* Evict the last not in use entry and use it */
496	Assert(!RTListIsEmpty(&pImage->ListLru));
497
498	RTListForEachReverse(&pImage->ListLru, pL2Entry, QEDL2CACHEENTRY, NodeLru)
499	{
500	if (!pL2Entry->cRefs)
501	break;
502	}
503
504	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
505	{
506	RTListNodeRemove(&pL2Entry->NodeSearch);
507	RTListNodeRemove(&pL2Entry->NodeLru);
508	pL2Entry->offL2Tbl = 0;
509	pL2Entry->cRefs = 1;
510	}
511	else
512	pL2Entry = NULL;
513	}
514
515	return pL2Entry;
516	}
517
518	/**
519	* Frees a L2 table cache entry.
520	*
521	* @returns nothing.
522	* @param pImage The image instance data.
523	* @param pL2Entry The L2 cache entry to free.
524	*/
525	static void qedL2TblCacheEntryFree(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
526	{
527	Assert(!pL2Entry->cRefs);
528	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
529	RTMemFree(pL2Entry);
530
531	pImage->cbL2Cache -= pImage->cbTable;
532	}
533
534	/**
535	* Inserts an entry in the L2 table cache.
536	*
537	* @returns nothing.
538	* @param pImage The image instance data.
539	* @param pL2Entry The L2 cache entry to insert.
540	*/
541	static void qedL2TblCacheEntryInsert(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
542	{
543	Assert(pL2Entry->offL2Tbl > 0);
544
545	/* Insert at the top of the LRU list. */
546	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
547
548	if (RTListIsEmpty(&pImage->ListSearch))
549	{
550	RTListAppend(&pImage->ListSearch, &pL2Entry->NodeSearch);
551	}
552	else
553	{
554	/* Insert into search list. */
555	PQEDL2CACHEENTRY pIt;
556	pIt = RTListGetFirst(&pImage->ListSearch, QEDL2CACHEENTRY, NodeSearch);
557	if (pIt->offL2Tbl > pL2Entry->offL2Tbl)
558	RTListPrepend(&pImage->ListSearch, &pL2Entry->NodeSearch);
559	else
560	{
561	bool fInserted = false;
562
563	RTListForEach(&pImage->ListSearch, pIt, QEDL2CACHEENTRY, NodeSearch)
564	{
565	Assert(pIt->offL2Tbl != pL2Entry->offL2Tbl);
566	if (pIt->offL2Tbl < pL2Entry->offL2Tbl)
567	{
568	RTListNodeInsertAfter(&pIt->NodeSearch, &pL2Entry->NodeSearch);
569	fInserted = true;
570	break;
571	}
572	}
573	Assert(fInserted);
574	}
575	}
576	}
577
578	/**
579	* Fetches the L2 from the given offset trying the LRU cache first and
580	* reading it from the image after a cache miss - version for async I/O.
581	*
582	* @returns VBox status code.
583	* @param pImage Image instance data.
584	* @param pIoCtx The I/O context.
585	* @param offL2Tbl The offset of the L2 table in the image.
586	* @param ppL2Entry Where to store the L2 table on success.
587	*/
588	static int qedL2TblCacheFetchAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
589	uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
590	{
591	int rc = VINF_SUCCESS;
592
593	/* Try to fetch the L2 table from the cache first. */
594	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
595	if (!pL2Entry)
596	{
597	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
598
599	if (pL2Entry)
600	{
601	/* Read from the image. */
602	PVDMETAXFER pMetaXfer;
603
604	pL2Entry->offL2Tbl = offL2Tbl;
605	rc = vdIfIoIntFileReadMeta(pImage->pIfIo, pImage->pStorage,
606	offL2Tbl, pL2Entry->paL2Tbl,
607	pImage->cbTable, pIoCtx,
608	&pMetaXfer, NULL, NULL);
609	if (RT_SUCCESS(rc))
610	{
611	vdIfIoIntMetaXferRelease(pImage->pIfIo, pMetaXfer);
612	#if defined(RT_BIG_ENDIAN)
613	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
614	#endif
615	qedL2TblCacheEntryInsert(pImage, pL2Entry);
616	}
617	else
618	{
619	qedL2TblCacheEntryRelease(pL2Entry);
620	qedL2TblCacheEntryFree(pImage, pL2Entry);
621	}
622	}
623	else
624	rc = VERR_NO_MEMORY;
625	}
626
627	if (RT_SUCCESS(rc))
628	*ppL2Entry = pL2Entry;
629
630	return rc;
631	}
632
633	/**
634	* Return power of 2 or 0 if num error.
635	*
636	* @returns The power of 2 or 0 if the given number is not a power of 2.
637	* @param u32 The number.
638	*/
639	static uint32_t qedGetPowerOfTwo(uint32_t u32)
640	{
641	if (u32 == 0)
642	return 0;
643	uint32_t uPower2 = 0;
644	while ((u32 & 1) == 0)
645	{
646	u32 >>= 1;
647	uPower2++;
648	}
649	return u32 == 1 ? uPower2 : 0;
650	}
651
652	/**
653	* Sets the L1, L2 and offset bitmasks and L1 and L2 bit shift members.
654	*
655	* @returns nothing.
656	* @param pImage The image instance data.
657	*/
658	static void qedTableMasksInit(PQEDIMAGE pImage)
659	{
660	uint32_t cClusterBits, cTableBits;
661
662	cClusterBits = qedGetPowerOfTwo(pImage->cbCluster);
663	cTableBits = qedGetPowerOfTwo(pImage->cTableEntries);
664
665	Assert(cClusterBits + 2 * cTableBits <= 64);
666
667	pImage->fOffsetMask = ((uint64_t)pImage->cbCluster - 1);
668	pImage->fL2Mask = ((uint64_t)pImage->cTableEntries - 1) << cClusterBits;
669	pImage->cL2Shift = cClusterBits;
670	pImage->fL1Mask = ((uint64_t)pImage->cTableEntries - 1) << (cClusterBits + cTableBits);
671	pImage->cL1Shift = cClusterBits + cTableBits;
672	}
673
674	/**
675	* Converts a given logical offset into the
676	*
677	* @returns nothing.
678	* @param pImage The image instance data.
679	* @param off The logical offset to convert.
680	* @param pidxL1 Where to store the index in the L1 table on success.
681	* @param pidxL2 Where to store the index in the L2 table on success.
682	* @param poffCluster Where to store the offset in the cluster on success.
683	*/
684	DECLINLINE(void) qedConvertLogicalOffset(PQEDIMAGE pImage, uint64_t off, uint32_t *pidxL1,
685	uint32_t pidxL2, uint32_t poffCluster)
686	{
687	AssertPtr(pidxL1);
688	AssertPtr(pidxL2);
689	AssertPtr(poffCluster);
690
691	*poffCluster = off & pImage->fOffsetMask;
692	*pidxL1 = (off & pImage->fL1Mask) >> pImage->cL1Shift;
693	*pidxL2 = (off & pImage->fL2Mask) >> pImage->cL2Shift;
694	}
695
696	/**
697	* Converts Cluster size to a byte size.
698	*
699	* @returns Number of bytes derived from the given number of clusters.
700	* @param pImage The image instance data.
701	* @param cClusters The clusters to convert.
702	*/
703	DECLINLINE(uint64_t) qedCluster2Byte(PQEDIMAGE pImage, uint64_t cClusters)
704	{
705	return cClusters * pImage->cbCluster;
706	}
707
708	/**
709	* Converts number of bytes to cluster size rounding to the next cluster.
710	*
711	* @returns Number of bytes derived from the given number of clusters.
712	* @param pImage The image instance data.
713	* @param cb Number of bytes to convert.
714	*/
715	DECLINLINE(uint64_t) qedByte2Cluster(PQEDIMAGE pImage, uint64_t cb)
716	{
717	return cb / pImage->cbCluster + (cb % pImage->cbCluster ? 1 : 0);
718	}
719
720	/**
721	* Allocates a new cluster in the image.
722	*
723	* @returns The start offset of the new cluster in the image.
724	* @param pImage The image instance data.
725	* @param cClusters Number of clusters to allocate.
726	*/
727	DECLINLINE(uint64_t) qedClusterAllocate(PQEDIMAGE pImage, uint32_t cClusters)
728	{
729	uint64_t offCluster;
730
731	offCluster = pImage->cbImage;
732	pImage->cbImage += cClusters*pImage->cbCluster;
733
734	return offCluster;
735	}
736
737	/**
738	* Returns the real image offset for a given cluster or an error if the cluster is not
739	* yet allocated.
740	*
741	* @returns VBox status code.
742	* VERR_VD_BLOCK_FREE if the cluster is not yet allocated.
743	* @param pImage The image instance data.
744	* @param pIoCtx The I/O context.
745	* @param idxL1 The L1 index.
746	* @param idxL2 The L2 index.
747	* @param offCluster Offset inside the cluster.
748	* @param poffImage Where to store the image offset on success;
749	*/
750	static int qedConvertToImageOffset(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
751	uint32_t idxL1, uint32_t idxL2,
752	uint32_t offCluster, uint64_t *poffImage)
753	{
754	int rc = VERR_VD_BLOCK_FREE;
755
756	AssertReturn(idxL1 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
757	AssertReturn(idxL2 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
758
759	if (pImage->paL1Table[idxL1])
760	{
761	PQEDL2CACHEENTRY pL2Entry;
762
763	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
764	&pL2Entry);
765	if (RT_SUCCESS(rc))
766	{
767	/* Get real file offset. */
768	if (pL2Entry->paL2Tbl[idxL2])
769	*poffImage = pL2Entry->paL2Tbl[idxL2] + offCluster;
770	else
771	rc = VERR_VD_BLOCK_FREE;
772
773	qedL2TblCacheEntryRelease(pL2Entry);
774	}
775	}
776
777	return rc;
778	}
779
780	/**
781	* Write the given table to image converting to the image endianess if required.
782	*
783	* @returns VBox status code.
784	* @param pImage The image instance data.
785	* @param pIoCtx The I/O context.
786	* @param offTbl The offset the table should be written to.
787	* @param paTbl The table to write.
788	* @param pfnComplete Callback called when the write completes.
789	* @param pvUser Opaque user data to pass in the completion callback.
790	*/
791	static int qedTblWrite(PQEDIMAGE pImage, PVDIOCTX pIoCtx, uint64_t offTbl, uint64_t *paTbl,
792	PFNVDXFERCOMPLETED pfnComplete, void *pvUser)
793	{
794	int rc = VINF_SUCCESS;
795
796	#if defined(RT_BIG_ENDIAN)
797	uint64_t paTblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
798	if (paTblImg)
799	{
800	qedTableConvertFromHostEndianess(paTblImg, paTbl,
801	pImage->cTableEntries);
802	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
803	offTbl, paTblImg, pImage->cbTable,
804	pIoCtx, pfnComplete, pvUser);
805	RTMemFree(paTblImg);
806	}
807	else
808	rc = VERR_NO_MEMORY;
809	#else
810	/* Write table directly. */
811	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
812	offTbl, paTbl, pImage->cbTable, pIoCtx,
813	pfnComplete, pvUser);
814	#endif
815
816	return rc;
817	}
818
819	/**
820	* Internal. Flush image data to disk.
821	*/
822	static int qedFlushImage(PQEDIMAGE pImage)
823	{
824	int rc = VINF_SUCCESS;
825
826	if ( pImage->pStorage
827	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
828	{
829	QedHeader Header;
830
831	Assert(!(pImage->cbTable % pImage->cbCluster));
832	#if defined(RT_BIG_ENDIAN)
833	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
834	if (paL1TblImg)
835	{
836	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
837	pImage->cTableEntries);
838	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
839	pImage->offL1Table, paL1TblImg,
840	pImage->cbTable);
841	RTMemFree(paL1TblImg);
842	}
843	else
844	rc = VERR_NO_MEMORY;
845	#else
846	/* Write L1 table directly. */
847	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, pImage->offL1Table,
848	pImage->paL1Table, pImage->cbTable);
849	#endif
850	if (RT_SUCCESS(rc))
851	{
852	/* Write header. */
853	qedHdrConvertFromHostEndianess(pImage, &Header);
854	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, 0, &Header,
855	sizeof(Header));
856	if (RT_SUCCESS(rc))
857	rc = vdIfIoIntFileFlushSync(pImage->pIfIo, pImage->pStorage);
858	}
859	}
860
861	return rc;
862	}
863
864	/**
865	* Checks whether the given cluster offset is valid.
866	*
867	* @returns Whether the given cluster offset is valid.
868	* @param offCluster The table offset to check.
869	* @param cbFile The real file size of the image.
870	* @param cbCluster The cluster size in bytes.
871	*/
872	DECLINLINE(bool) qedIsClusterOffsetValid(uint64_t offCluster, uint64_t cbFile, size_t cbCluster)
873	{
874	return (offCluster <= cbFile - cbCluster)
875	&& !(offCluster & (cbCluster - 1));
876	}
877
878	/**
879	* Checks whether the given table offset is valid.
880	*
881	* @returns Whether the given table offset is valid.
882	* @param offTbl The table offset to check.
883	* @param cbFile The real file size of the image.
884	* @param cbTable The table size in bytes.
885	* @param cbCluster The cluster size in bytes.
886	*/
887	DECLINLINE(bool) qedIsTblOffsetValid(uint64_t offTbl, uint64_t cbFile, size_t cbTable, size_t cbCluster)
888	{
889	return (offTbl <= cbFile - cbTable)
890	&& !(offTbl & (cbCluster - 1));
891	}
892
893	/**
894	* Sets the specified range in the cluster bitmap checking whether any of the clusters is already
895	* used before.
896	*
897	* @returns Whether the range was clear and is set now.
898	* @param pvClusterBitmap The cluster bitmap to use.
899	* @param offClusterStart The first cluster to check and set.
900	* @param offClusterEnd The first cluster to not check and set anymore.
901	*/
902	static bool qedClusterBitmapCheckAndSet(void *pvClusterBitmap, uint32_t offClusterStart, uint32_t offClusterEnd)
903	{
904	for (uint32_t offCluster = offClusterStart; offCluster < offClusterEnd; offCluster++)
905	if (ASMBitTest(pvClusterBitmap, offCluster))
906	return false;
907
908	ASMBitSetRange(pvClusterBitmap, offClusterStart, offClusterEnd);
909	return true;
910	}
911
912	/**
913	* Checks the given image for consistency, usually called when the
914	* QED_FEATURE_NEED_CHECK bit is set.
915	*
916	* @returns VBox status code.
917	* @retval VINF_SUCCESS when the image can be accessed.
918	* @param pImage The image instance data.
919	* @param pHeader The header to use for checking.
920	*
921	* @note It is not required that the image state is fully initialized Only
922	* The I/O interface and storage handle need to be valid.
923	* @note The header must be converted to the host CPU endian format already
924	* and should be validated already.
925	*/
926	static int qedCheckImage(PQEDIMAGE pImage, PQedHeader pHeader)
927	{
928	uint64_t cbFile;
929	uint32_t cbTable;
930	uint32_t cTableEntries;
931	uint64_t *paL1Tbl = NULL;
932	uint64_t *paL2Tbl = NULL;
933	void *pvClusterBitmap = NULL;
934	uint32_t offClusterStart;
935	int rc = VINF_SUCCESS;
936
937	pImage->cbCluster = pHeader->u32ClusterSize;
938	cbTable = pHeader->u32TableSize * pHeader->u32ClusterSize;
939	cTableEntries = cbTable / sizeof(uint64_t);
940
941	do
942	{
943	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
944	if (RT_FAILURE(rc))
945	{
946	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
947	N_("Qed: Querying the file size of image '%s' failed"),
948	pImage->pszFilename);
949	break;
950	}
951
952	/* Allocate L1 table. */
953	paL1Tbl = (uint64_t *)RTMemAllocZ(cbTable);
954	if (!paL1Tbl)
955	{
956	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
957	N_("Qed: Allocating memory for the L1 table for image '%s' failed"),
958	pImage->pszFilename);
959	break;
960	}
961
962	paL2Tbl = (uint64_t *)RTMemAllocZ(cbTable);
963	if (!paL2Tbl)
964	{
965	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
966	N_("Qed: Allocating memory for the L2 table for image '%s' failed"),
967	pImage->pszFilename);
968	break;
969	}
970
971	pvClusterBitmap = RTMemAllocZ(cbFile / pHeader->u32ClusterSize / 8);
972	if (!pvClusterBitmap)
973	{
974	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
975	N_("Qed: Allocating memory for the cluster bitmap for image '%s' failed"),
976	pImage->pszFilename);
977	break;
978	}
979
980	/* Validate L1 table offset. */
981	if (!qedIsTblOffsetValid(pHeader->u64OffL1Table, cbFile, cbTable, pHeader->u32ClusterSize))
982	{
983	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
984	N_("Qed: L1 table offset of image '%s' is corrupt (%llu)"),
985	pImage->pszFilename, pHeader->u64OffL1Table);
986	break;
987	}
988
989	/* Read L1 table. */
990	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
991	pHeader->u64OffL1Table, paL1Tbl, cbTable);
992	if (RT_FAILURE(rc))
993	{
994	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
995	N_("Qed: Reading the L1 table from image '%s' failed"),
996	pImage->pszFilename);
997	break;
998	}
999
1000	/* Mark the L1 table in cluster bitmap. */
1001	ASMBitSet(pvClusterBitmap, 0); /* Header is always in cluster 0. */
1002	offClusterStart = qedByte2Cluster(pImage, pHeader->u64OffL1Table);
1003	bool fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1004	Assert(fSet);
1005
1006	/* Scan the L1 and L2 tables for invalid entries. */
1007	qedTableConvertToHostEndianess(paL1Tbl, cTableEntries);
1008
1009	for (unsigned iL1 = 0; iL1 < cTableEntries; iL1++)
1010	{
1011	if (!paL1Tbl[iL1])
1012	continue; /* Skip unallocated clusters. */
1013
1014	if (!qedIsTblOffsetValid(paL1Tbl[iL1], cbFile, cbTable, pHeader->u32ClusterSize))
1015	{
1016	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1017	N_("Qed: Entry %d of the L1 table from image '%s' is invalid (%llu)"),
1018	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1019	break;
1020	}
1021
1022	/* Now check that the clusters are not allocated already. */
1023	offClusterStart = qedByte2Cluster(pImage, paL1Tbl[iL1]);
1024	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1025	if (!fSet)
1026	{
1027	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1028	N_("Qed: Entry %d of the L1 table from image '%s' points to a already used cluster (%llu)"),
1029	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1030	break;
1031	}
1032
1033	/* Read the linked L2 table and check it. */
1034	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1035	paL1Tbl[iL1], paL2Tbl, cbTable);
1036	if (RT_FAILURE(rc))
1037	{
1038	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1039	N_("Qed: Reading the L2 table from image '%s' failed"),
1040	pImage->pszFilename);
1041	break;
1042	}
1043
1044	/* Check all L2 entries. */
1045	for (unsigned iL2 = 0; iL2 < cTableEntries; iL2++)
1046	{
1047	if (paL2Tbl[iL2])
1048	continue; /* Skip unallocated clusters. */
1049
1050	if (!qedIsClusterOffsetValid(paL2Tbl[iL2], cbFile, pHeader->u32ClusterSize))
1051	{
1052	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1053	N_("Qed: Entry %d of the L2 table from image '%s' is invalid (%llu)"),
1054	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1055	break;
1056	}
1057
1058	/* Now check that the clusters are not allocated already. */
1059	offClusterStart = qedByte2Cluster(pImage, paL2Tbl[iL2]);
1060	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + 1);
1061	if (!fSet)
1062	{
1063	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1064	N_("Qed: Entry %d of the L2 table from image '%s' points to a already used cluster (%llu)"),
1065	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1066	break;
1067	}
1068	}
1069	}
1070	} while(0);
1071
1072	if (paL1Tbl)
1073	RTMemFree(paL1Tbl);
1074	if (paL2Tbl)
1075	RTMemFree(paL2Tbl);
1076	if (pvClusterBitmap)
1077	RTMemFree(pvClusterBitmap);
1078
1079	return rc;
1080	}
1081
1082	/**
1083	* Internal. Free all allocated space for representing an image except pImage,
1084	* and optionally delete the image from disk.
1085	*/
1086	static int qedFreeImage(PQEDIMAGE pImage, bool fDelete)
1087	{
1088	int rc = VINF_SUCCESS;
1089
1090	/* Freeing a never allocated image (e.g. because the open failed) is
1091	* not signalled as an error. After all nothing bad happens. */
1092	if (pImage)
1093	{
1094	if (pImage->pStorage)
1095	{
1096	/* No point updating the file that is deleted anyway. */
1097	if (!fDelete)
1098	qedFlushImage(pImage);
1099
1100	rc = vdIfIoIntFileClose(pImage->pIfIo, pImage->pStorage);
1101	pImage->pStorage = NULL;
1102	}
1103
1104	if (pImage->paL1Table)
1105	RTMemFree(pImage->paL1Table);
1106
1107	if (pImage->pszBackingFilename)
1108	{
1109	RTMemFree(pImage->pszBackingFilename);
1110	pImage->pszBackingFilename = NULL;
1111	}
1112
1113	qedL2TblCacheDestroy(pImage);
1114
1115	if (fDelete && pImage->pszFilename)
1116	vdIfIoIntFileDelete(pImage->pIfIo, pImage->pszFilename);
1117	}
1118
1119	LogFlowFunc(("returns %Rrc\n", rc));
1120	return rc;
1121	}
1122
1123	/**
1124	* Internal: Open an image, constructing all necessary data structures.
1125	*/
1126	static int qedOpenImage(PQEDIMAGE pImage, unsigned uOpenFlags)
1127	{
1128	pImage->uOpenFlags = uOpenFlags;
1129
1130	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1131	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1132	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1133
1134	/*
1135	* Create the L2 cache before opening the image so we can call qedFreeImage()
1136	* even if opening the image file fails.
1137	*/
1138	int rc = qedL2TblCacheCreate(pImage);
1139	if (RT_SUCCESS(rc))
1140	{
1141	/* Open the image. */
1142	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename,
1143	VDOpenFlagsToFileOpenFlags(uOpenFlags,
1144	false /* fCreate */),
1145	&pImage->pStorage);
1146	if (RT_SUCCESS(rc))
1147	{
1148	uint64_t cbFile;
1149	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1150	if ( RT_SUCCESS(rc)
1151	&& cbFile > sizeof(QedHeader))
1152	{
1153	QedHeader Header;
1154
1155	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, 0, &Header, sizeof(Header));
1156	if ( RT_SUCCESS(rc)
1157	&& qedHdrConvertToHostEndianess(&Header))
1158	{
1159	if ( !(Header.u64FeatureFlags & ~QED_FEATURE_MASK)
1160	&& !(Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE_NO_PROBE))
1161	{
1162	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1163	{
1164	/* Image needs checking. */
1165	if (!(uOpenFlags & VD_OPEN_FLAGS_READONLY))
1166	rc = qedCheckImage(pImage, &Header);
1167	else
1168	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1169	N_("Qed: Image '%s' needs checking but is opened readonly"),
1170	pImage->pszFilename);
1171	}
1172
1173	if ( RT_SUCCESS(rc)
1174	&& (Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE))
1175	{
1176	/* Load backing filename from image. */
1177	pImage->pszBackingFilename = (char )RTMemAllocZ(Header.u32BackingFilenameSize + 1); / +1 for \0 terminator. */
1178	if (pImage->pszBackingFilename)
1179	{
1180	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1181	pImage->offBackingFilename = Header.u32OffBackingFilename;
1182	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1183	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1184	Header.u32BackingFilenameSize);
1185	}
1186	else
1187	rc = VERR_NO_MEMORY;
1188	}
1189
1190	if (RT_SUCCESS(rc))
1191	{
1192	pImage->cbImage = cbFile;
1193	pImage->cbCluster = Header.u32ClusterSize;
1194	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1195	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1196	pImage->offL1Table = Header.u64OffL1Table;
1197	pImage->cbSize = Header.u64Size;
1198	qedTableMasksInit(pImage);
1199
1200	/* Allocate L1 table. */
1201	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1202	if (pImage->paL1Table)
1203	{
1204	/* Read from the image. */
1205	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1206	pImage->offL1Table, pImage->paL1Table,
1207	pImage->cbTable);
1208	if (RT_SUCCESS(rc))
1209	{
1210	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1211
1212	/* If the consistency check succeeded, clear the flag by flushing the image. */
1213	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1214	rc = qedFlushImage(pImage);
1215	}
1216	else
1217	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1218	N_("Qed: Reading the L1 table for image '%s' failed"),
1219	pImage->pszFilename);
1220	}
1221	else
1222	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1223	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1224	pImage->pszFilename);
1225	}
1226	}
1227	else
1228	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1229	N_("Qed: The image '%s' makes use of unsupported features"),
1230	pImage->pszFilename);
1231	}
1232	else if (RT_SUCCESS(rc))
1233	rc = VERR_VD_GEN_INVALID_HEADER;
1234	}
1235	else if (RT_SUCCESS(rc))
1236	rc = VERR_VD_GEN_INVALID_HEADER;
1237	}
1238	/* else: Do NOT signal an appropriate error here, as the VD layer has the
1239	* choice of retrying the open if it failed. */
1240	}
1241	else
1242	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1243	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1244	pImage->pszFilename);
1245
1246	if (RT_SUCCESS(rc))
1247	{
1248	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1249	pImage->RegionList.fFlags = 0;
1250	pImage->RegionList.cRegions = 1;
1251
1252	pRegion->offRegion = 0; /* Disk start. */
1253	pRegion->cbBlock = 512;
1254	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1255	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1256	pRegion->cbData = 512;
1257	pRegion->cbMetadata = 0;
1258	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1259	}
1260	else
1261	qedFreeImage(pImage, false);
1262	return rc;
1263	}
1264
1265	/**
1266	* Internal: Create a qed image.
1267	*/
1268	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1269	unsigned uImageFlags, const char *pszComment,
1270	PCVDGEOMETRY pPCHSGeometry,
1271	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1272	PVDINTERFACEPROGRESS pIfProgress,
1273	unsigned uPercentStart, unsigned uPercentSpan)
1274	{
1275	RT_NOREF1(pszComment);
1276	int rc;
1277
1278	if (!(uImageFlags & VD_IMAGE_FLAGS_FIXED))
1279	{
1280	rc = qedL2TblCacheCreate(pImage);
1281	if (RT_SUCCESS(rc))
1282	{
1283	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1284	pImage->uImageFlags = uImageFlags;
1285	pImage->PCHSGeometry = *pPCHSGeometry;
1286	pImage->LCHSGeometry = *pLCHSGeometry;
1287
1288	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1289	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1290	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1291
1292	/* Create image file. */
1293	uint32_t fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1294	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1295	if (RT_SUCCESS(rc))
1296	{
1297	/* Init image state. */
1298	pImage->cbSize = cbSize;
1299	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1300	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1301	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1302	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1303	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1304	pImage->cbBackingFilename = 0;
1305	pImage->offBackingFilename = 0;
1306	qedTableMasksInit(pImage);
1307
1308	/* Init L1 table. */
1309	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1310	if (RT_LIKELY(pImage->paL1Table))
1311	{
1312	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan * 98 / 100);
1313	rc = qedFlushImage(pImage);
1314	}
1315	else
1316	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1317	pImage->pszFilename);
1318	}
1319	else
1320	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1321	}
1322	else
1323	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1324	pImage->pszFilename);
1325	}
1326	else
1327	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1328
1329	if (RT_SUCCESS(rc))
1330	{
1331	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1332	pImage->RegionList.fFlags = 0;
1333	pImage->RegionList.cRegions = 1;
1334
1335	pRegion->offRegion = 0; /* Disk start. */
1336	pRegion->cbBlock = 512;
1337	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1338	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1339	pRegion->cbData = 512;
1340	pRegion->cbMetadata = 0;
1341	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1342
1343	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan);
1344	}
1345	else
1346	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1347
1348	return rc;
1349	}
1350
1351	/**
1352	* Rollback anything done during async cluster allocation.
1353	*
1354	* @returns VBox status code.
1355	* @param pImage The image instance data.
1356	* @param pIoCtx The I/O context.
1357	* @param pClusterAlloc The cluster allocation to rollback.
1358	*/
1359	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1360	{
1361	RT_NOREF1(pIoCtx);
1362	int rc = VINF_SUCCESS;
1363
1364	switch (pClusterAlloc->enmAllocState)
1365	{
1366	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1367	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1368	{
1369	/* Revert the L1 table entry */
1370	pImage->paL1Table[pClusterAlloc->idxL1] = 0;
1371	pImage->pL2TblAlloc = NULL;
1372
1373	/* Assumption right now is that the L1 table is not modified on storage if the link fails. */
1374	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1375	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1376	Assert(!pClusterAlloc->pL2Entry->cRefs);
1377	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1378	break;
1379	}
1380	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1381	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1382	{
1383	/* Assumption right now is that the L2 table is not modified if the link fails. */
1384	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = 0;
1385	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1386	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1387	break;
1388	}
1389	default:
1390	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1391	rc = VERR_INVALID_STATE;
1392	}
1393
1394	RTMemFree(pClusterAlloc);
1395	return rc;
1396	}
1397
1398	/**
1399	* Updates the state of the async cluster allocation.
1400	*
1401	* @returns VBox status code.
1402	* @param pBackendData The opaque backend data.
1403	* @param pIoCtx I/O context associated with this request.
1404	* @param pvUser Opaque user data passed during a read/write request.
1405	* @param rcReq Status code for the completed request.
1406	*/
1407	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1408	{
1409	int rc = VINF_SUCCESS;
1410	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1411	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1412
1413	if (RT_FAILURE(rcReq))
1414	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1415
1416	AssertPtr(pClusterAlloc->pL2Entry);
1417
1418	switch (pClusterAlloc->enmAllocState)
1419	{
1420	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1421	{
1422	/* Update the link in the in memory L1 table now. */
1423	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1424
1425	/* Update the link in the on disk L1 table now. */
1426	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1427	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1428	qedAsyncClusterAllocUpdate, pClusterAlloc);
1429	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1430	break;
1431	else if (RT_FAILURE(rc))
1432	{
1433	/* Rollback. */
1434	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1435	break;
1436	}
1437	}
1438	RT_FALL_THRU();
1439	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1440	{
1441	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1442	uint64_t offData = qedClusterAllocate(pImage, 1);
1443
1444	pImage->pL2TblAlloc = NULL;
1445	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1446
1447	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1448	pClusterAlloc->cbImageOld = offData;
1449	pClusterAlloc->offClusterNew = offData;
1450
1451	/* Write data. */
1452	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1453	offData, pIoCtx, pClusterAlloc->cbToWrite,
1454	qedAsyncClusterAllocUpdate, pClusterAlloc);
1455	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1456	break;
1457	else if (RT_FAILURE(rc))
1458	{
1459	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1460	RTMemFree(pClusterAlloc);
1461	break;
1462	}
1463	}
1464	RT_FALL_THRU();
1465	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1466	{
1467	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1468	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1469
1470	/* Link L2 table and update it. */
1471	rc = qedTblWrite(pImage, pIoCtx, pImage->paL1Table[pClusterAlloc->idxL1],
1472	pClusterAlloc->pL2Entry->paL2Tbl,
1473	qedAsyncClusterAllocUpdate, pClusterAlloc);
1474	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1475	break;
1476	else if (RT_FAILURE(rc))
1477	{
1478	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1479	RTMemFree(pClusterAlloc);
1480	break;
1481	}
1482	}
1483	RT_FALL_THRU();
1484	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1485	{
1486	/* Everything done without errors, signal completion. */
1487	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1488	RTMemFree(pClusterAlloc);
1489	rc = VINF_SUCCESS;
1490	break;
1491	}
1492	default:
1493	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1494	pClusterAlloc->enmAllocState));
1495	}
1496
1497	return rc;
1498	}
1499
1500	/** @copydoc VDIMAGEBACKEND::pfnProbe */
1501	static DECLCALLBACK(int) qedProbe(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1502	PVDINTERFACE pVDIfsImage, VDTYPE *penmType)
1503	{
1504	RT_NOREF1(pVDIfsDisk);
1505	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1506	PVDIOSTORAGE pStorage = NULL;
1507	int rc = VINF_SUCCESS;
1508
1509	/* Get I/O interface. */
1510	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1511	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1512	AssertReturn((VALID_PTR(pszFilename) && *pszFilename), VERR_INVALID_PARAMETER);
1513
1514	/*
1515	* Open the file and read the footer.
1516	*/
1517	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1518	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1519	false /* fCreate */),
1520	&pStorage);
1521	if (RT_SUCCESS(rc))
1522	{
1523	uint64_t cbFile;
1524
1525	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1526	if ( RT_SUCCESS(rc)
1527	&& cbFile > sizeof(QedHeader))
1528	{
1529	QedHeader Header;
1530
1531	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1532	if ( RT_SUCCESS(rc)
1533	&& qedHdrConvertToHostEndianess(&Header))
1534	*penmType = VDTYPE_HDD;
1535	else
1536	rc = VERR_VD_GEN_INVALID_HEADER;
1537	}
1538	else
1539	rc = VERR_VD_GEN_INVALID_HEADER;
1540	}
1541
1542	if (pStorage)
1543	vdIfIoIntFileClose(pIfIo, pStorage);
1544
1545	LogFlowFunc(("returns %Rrc\n", rc));
1546	return rc;
1547	}
1548
1549	/** @copydoc VDIMAGEBACKEND::pfnOpen */
1550	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1551	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1552	VDTYPE enmType, void **ppBackendData)
1553	{
1554	RT_NOREF1(enmType); /*< @todo r=klaus make use of the type info. /
1555
1556	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n",
1557	pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1558	int rc;
1559
1560	/* Check open flags. All valid flags are supported. */
1561	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1562	AssertReturn((VALID_PTR(pszFilename) && *pszFilename), VERR_INVALID_PARAMETER);
1563
1564	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1565	if (RT_LIKELY(pImage))
1566	{
1567	pImage->pszFilename = pszFilename;
1568	pImage->pStorage = NULL;
1569	pImage->pVDIfsDisk = pVDIfsDisk;
1570	pImage->pVDIfsImage = pVDIfsImage;
1571
1572	rc = qedOpenImage(pImage, uOpenFlags);
1573	if (RT_SUCCESS(rc))
1574	*ppBackendData = pImage;
1575	else
1576	RTMemFree(pImage);
1577	}
1578	else
1579	rc = VERR_NO_MEMORY;
1580
1581	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1582	return rc;
1583	}
1584
1585	/** @copydoc VDIMAGEBACKEND::pfnCreate */
1586	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1587	unsigned uImageFlags, const char *pszComment,
1588	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1589	PCRTUUID pUuid, unsigned uOpenFlags,
1590	unsigned uPercentStart, unsigned uPercentSpan,
1591	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1592	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1593	void **ppBackendData)
1594	{
1595	RT_NOREF1(pUuid);
1596	LogFlowFunc(("pszFilename=\"%s\" cbSize=%llu uImageFlags=%#x pszComment=\"%s\" pPCHSGeometry=%#p pLCHSGeometry=%#p Uuid=%RTuuid uOpenFlags=%#x uPercentStart=%u uPercentSpan=%u pVDIfsDisk=%#p pVDIfsImage=%#p pVDIfsOperation=%#p enmType=%d ppBackendData=%#p",
1597	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1598	int rc;
1599
1600	/* Check the VD container type. */
1601	if (enmType != VDTYPE_HDD)
1602	return VERR_VD_INVALID_TYPE;
1603
1604	/* Check open flags. All valid flags are supported. */
1605	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1606	AssertReturn( VALID_PTR(pszFilename)
1607	&& *pszFilename
1608	&& VALID_PTR(pPCHSGeometry)
1609	&& VALID_PTR(pLCHSGeometry), VERR_INVALID_PARAMETER);
1610
1611	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1612	if (RT_LIKELY(pImage))
1613	{
1614	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1615
1616	pImage->pszFilename = pszFilename;
1617	pImage->pStorage = NULL;
1618	pImage->pVDIfsDisk = pVDIfsDisk;
1619	pImage->pVDIfsImage = pVDIfsImage;
1620
1621	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1622	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1623	pIfProgress, uPercentStart, uPercentSpan);
1624	if (RT_SUCCESS(rc))
1625	{
1626	/* So far the image is opened in read/write mode. Make sure the
1627	* image is opened in read-only mode if the caller requested that. */
1628	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1629	{
1630	qedFreeImage(pImage, false);
1631	rc = qedOpenImage(pImage, uOpenFlags);
1632	}
1633
1634	if (RT_SUCCESS(rc))
1635	*ppBackendData = pImage;
1636	}
1637
1638	if (RT_FAILURE(rc))
1639	RTMemFree(pImage);
1640	}
1641	else
1642	rc = VERR_NO_MEMORY;
1643
1644	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1645	return rc;
1646	}
1647
1648	/** @copydoc VDIMAGEBACKEND::pfnRename */
1649	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1650	{
1651	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1652	int rc = VINF_SUCCESS;
1653	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1654
1655	/* Check arguments. */
1656	AssertReturn((pImage && pszFilename && *pszFilename), VERR_INVALID_PARAMETER);
1657
1658	/* Close the image. */
1659	rc = qedFreeImage(pImage, false);
1660	if (RT_SUCCESS(rc))
1661	{
1662	/* Rename the file. */
1663	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1664	if (RT_SUCCESS(rc))
1665	{
1666	/* Update pImage with the new information. */
1667	pImage->pszFilename = pszFilename;
1668
1669	/* Open the old image with new name. */
1670	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1671	}
1672	else
1673	{
1674	/* The move failed, try to reopen the original image. */
1675	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1676	if (RT_FAILURE(rc2))
1677	rc = rc2;
1678	}
1679	}
1680
1681	LogFlowFunc(("returns %Rrc\n", rc));
1682	return rc;
1683	}
1684
1685	/** @copydoc VDIMAGEBACKEND::pfnClose */
1686	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1687	{
1688	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1689	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1690
1691	int rc = qedFreeImage(pImage, fDelete);
1692	RTMemFree(pImage);
1693
1694	LogFlowFunc(("returns %Rrc\n", rc));
1695	return rc;
1696	}
1697
1698	/** @copydoc VDIMAGEBACKEND::pfnRead */
1699	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1700	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1701	{
1702	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1703	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1704	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1705	uint32_t offCluster = 0;
1706	uint32_t idxL1 = 0;
1707	uint32_t idxL2 = 0;
1708	uint64_t offFile = 0;
1709
1710	AssertPtr(pImage);
1711	Assert(uOffset % 512 == 0);
1712	Assert(cbToRead % 512 == 0);
1713	AssertReturn((VALID_PTR(pIoCtx) && cbToRead), VERR_INVALID_PARAMETER);
1714	AssertReturn(uOffset + cbToRead <= pImage->cbSize, VERR_INVALID_PARAMETER);
1715
1716	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1717
1718	/* Clip read size to remain in the cluster. */
1719	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1720
1721	/* Get offset in image. */
1722	int rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1723	if (RT_SUCCESS(rc))
1724	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1725	pIoCtx, cbToRead);
1726
1727	if ( ( RT_SUCCESS(rc)
1728	\|\| rc == VERR_VD_BLOCK_FREE
1729	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1730	&& pcbActuallyRead)
1731	*pcbActuallyRead = cbToRead;
1732
1733	LogFlowFunc(("returns %Rrc\n", rc));
1734	return rc;
1735	}
1736
1737	/** @copydoc VDIMAGEBACKEND::pfnWrite */
1738	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1739	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1740	size_t *pcbPostRead, unsigned fWrite)
1741	{
1742	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1743	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1744	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1745	uint32_t offCluster = 0;
1746	uint32_t idxL1 = 0;
1747	uint32_t idxL2 = 0;
1748	uint64_t offImage = 0;
1749	int rc = VINF_SUCCESS;
1750
1751	AssertPtr(pImage);
1752	Assert(!(uOffset % 512));
1753	Assert(!(cbToWrite % 512));
1754	AssertReturn((VALID_PTR(pIoCtx) && cbToWrite), VERR_INVALID_PARAMETER);
1755	AssertReturn(uOffset + cbToWrite <= pImage->cbSize, VERR_INVALID_PARAMETER);
1756
1757	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1758	{
1759	/* Convert offset to L1, L2 index and cluster offset. */
1760	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1761
1762	/* Clip write size to remain in the cluster. */
1763	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1764	Assert(!(cbToWrite % 512));
1765
1766	/* Get offset in image. */
1767	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1768	if (RT_SUCCESS(rc))
1769	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1770	offImage, pIoCtx, cbToWrite, NULL, NULL);
1771	else if (rc == VERR_VD_BLOCK_FREE)
1772	{
1773	if ( cbToWrite == pImage->cbCluster
1774	&& !(fWrite & VD_WRITE_NO_ALLOC))
1775	{
1776	PQEDL2CACHEENTRY pL2Entry = NULL;
1777
1778	/* Full cluster write to previously unallocated cluster.
1779	* Allocate cluster and write data. */
1780	Assert(!offCluster);
1781
1782	do
1783	{
1784	/* Check if we have to allocate a new cluster for L2 tables. */
1785	if (!pImage->paL1Table[idxL1])
1786	{
1787	uint64_t offL2Tbl;
1788	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1789
1790	/* Allocate new async cluster allocation state. */
1791	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1792	if (RT_UNLIKELY(!pL2ClusterAlloc))
1793	{
1794	rc = VERR_NO_MEMORY;
1795	break;
1796	}
1797
1798	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1799	if (!pL2Entry)
1800	{
1801	rc = VERR_NO_MEMORY;
1802	RTMemFree(pL2ClusterAlloc);
1803	break;
1804	}
1805
1806	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1807	pL2Entry->offL2Tbl = offL2Tbl;
1808	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1809
1810	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1811	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1812	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1813	pL2ClusterAlloc->idxL1 = idxL1;
1814	pL2ClusterAlloc->idxL2 = idxL2;
1815	pL2ClusterAlloc->cbToWrite = cbToWrite;
1816	pL2ClusterAlloc->pL2Entry = pL2Entry;
1817
1818	pImage->pL2TblAlloc = pL2Entry;
1819
1820	LogFlowFunc(("Allocating new L2 table at cluster offset %llu\n", offL2Tbl));
1821
1822	/*
1823	* Write the L2 table first and link to the L1 table afterwards.
1824	* If something unexpected happens the worst case which can happen
1825	* is a leak of some clusters.
1826	*/
1827	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1828	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1829	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1830	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1831	break;
1832	else if (RT_FAILURE(rc))
1833	{
1834	RTMemFree(pL2ClusterAlloc);
1835	qedL2TblCacheEntryFree(pImage, pL2Entry);
1836	break;
1837	}
1838
1839	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1840	}
1841	else
1842	{
1843	LogFlowFunc(("Fetching L2 table at cluster offset %llu\n", pImage->paL1Table[idxL1]));
1844
1845	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1846	&pL2Entry);
1847
1848	if (RT_SUCCESS(rc))
1849	{
1850	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1851
1852	/* Allocate new async cluster allocation state. */
1853	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1854	if (RT_UNLIKELY(!pDataClusterAlloc))
1855	{
1856	rc = VERR_NO_MEMORY;
1857	break;
1858	}
1859
1860	/* Allocate new cluster for the data. */
1861	uint64_t offData = qedClusterAllocate(pImage, 1);
1862
1863	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1864	pDataClusterAlloc->cbImageOld = offData;
1865	pDataClusterAlloc->offClusterNew = offData;
1866	pDataClusterAlloc->idxL1 = idxL1;
1867	pDataClusterAlloc->idxL2 = idxL2;
1868	pDataClusterAlloc->cbToWrite = cbToWrite;
1869	pDataClusterAlloc->pL2Entry = pL2Entry;
1870
1871	/* Write data. */
1872	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1873	offData, pIoCtx, cbToWrite,
1874	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1875	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1876	break;
1877	else if (RT_FAILURE(rc))
1878	{
1879	RTMemFree(pDataClusterAlloc);
1880	break;
1881	}
1882
1883	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1884	}
1885	}
1886
1887	} while (0);
1888
1889	*pcbPreRead = 0;
1890	*pcbPostRead = 0;
1891	}
1892	else
1893	{
1894	/* Trying to do a partial write to an unallocated cluster. Don't do
1895	* anything except letting the upper layer know what to do. */
1896	*pcbPreRead = offCluster;
1897	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
1898	}
1899	}
1900
1901	if (pcbWriteProcess)
1902	*pcbWriteProcess = cbToWrite;
1903	}
1904	else
1905	rc = VERR_VD_IMAGE_READ_ONLY;
1906
1907	LogFlowFunc(("returns %Rrc\n", rc));
1908	return rc;
1909	}
1910
1911	/** @copydoc VDIMAGEBACKEND::pfnFlush */
1912	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
1913	{
1914	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1915	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1916	int rc = VINF_SUCCESS;
1917
1918	AssertPtr(pImage);
1919	AssertPtrReturn(pIoCtx, VERR_INVALID_PARAMETER);
1920
1921	if ( pImage->pStorage
1922	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1923	{
1924	QedHeader Header;
1925
1926	Assert(!(pImage->cbTable % pImage->cbCluster));
1927	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1928	NULL, NULL);
1929	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1930	{
1931	/* Write header. */
1932	qedHdrConvertFromHostEndianess(pImage, &Header);
1933	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1934	0, &Header, sizeof(Header),
1935	pIoCtx, NULL, NULL);
1936	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1937	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
1938	pIoCtx, NULL, NULL);
1939	}
1940	}
1941
1942	LogFlowFunc(("returns %Rrc\n", rc));
1943	return rc;
1944	}
1945
1946	/** @copydoc VDIMAGEBACKEND::pfnGetVersion */
1947	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
1948	{
1949	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1950	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1951
1952	AssertPtrReturn(pImage, 0);
1953
1954	return 1;
1955	}
1956
1957	/** @copydoc VDIMAGEBACKEND::pfnGetFileSize */
1958	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
1959	{
1960	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1961	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1962	uint64_t cb = 0;
1963
1964	AssertPtrReturn(pImage, 0);
1965
1966	uint64_t cbFile;
1967	if (pImage->pStorage)
1968	{
1969	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1970	if (RT_SUCCESS(rc))
1971	cb += cbFile;
1972	}
1973
1974	LogFlowFunc(("returns %lld\n", cb));
1975	return cb;
1976	}
1977
1978	/** @copydoc VDIMAGEBACKEND::pfnGetPCHSGeometry */
1979	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
1980	PVDGEOMETRY pPCHSGeometry)
1981	{
1982	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
1983	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1984	int rc = VINF_SUCCESS;
1985
1986	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
1987
1988	if (pImage->PCHSGeometry.cCylinders)
1989	*pPCHSGeometry = pImage->PCHSGeometry;
1990	else
1991	rc = VERR_VD_GEOMETRY_NOT_SET;
1992
1993	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
1994	return rc;
1995	}
1996
1997	/** @copydoc VDIMAGEBACKEND::pfnSetPCHSGeometry */
1998	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
1999	PCVDGEOMETRY pPCHSGeometry)
2000	{
2001	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n",
2002	pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2003	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2004	int rc = VINF_SUCCESS;
2005
2006	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2007
2008	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2009	rc = VERR_VD_IMAGE_READ_ONLY;
2010	else
2011	pImage->PCHSGeometry = *pPCHSGeometry;
2012
2013	LogFlowFunc(("returns %Rrc\n", rc));
2014	return rc;
2015	}
2016
2017	/** @copydoc VDIMAGEBACKEND::pfnGetLCHSGeometry */
2018	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData, PVDGEOMETRY pLCHSGeometry)
2019	{
2020	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2021	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2022	int rc = VINF_SUCCESS;
2023
2024	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2025
2026	if (pImage->LCHSGeometry.cCylinders)
2027	*pLCHSGeometry = pImage->LCHSGeometry;
2028	else
2029	rc = VERR_VD_GEOMETRY_NOT_SET;
2030
2031	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders,
2032	pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2033	return rc;
2034	}
2035
2036	/** @copydoc VDIMAGEBACKEND::pfnSetLCHSGeometry */
2037	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData, PCVDGEOMETRY pLCHSGeometry)
2038	{
2039	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData,
2040	pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2041	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2042	int rc = VINF_SUCCESS;
2043
2044	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2045
2046	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2047	rc = VERR_VD_IMAGE_READ_ONLY;
2048	else
2049	pImage->LCHSGeometry = *pLCHSGeometry;
2050
2051	LogFlowFunc(("returns %Rrc\n", rc));
2052	return rc;
2053	}
2054
2055	/** @copydoc VDIMAGEBACKEND::pfnQueryRegions */
2056	static DECLCALLBACK(int) qedQueryRegions(void pBackendData, PCVDREGIONLIST ppRegionList)
2057	{
2058	LogFlowFunc(("pBackendData=%#p ppRegionList=%#p\n", pBackendData, ppRegionList));
2059	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2060
2061	AssertPtrReturn(pThis, VERR_VD_NOT_OPENED);
2062
2063	*ppRegionList = &pThis->RegionList;
2064	LogFlowFunc(("returns %Rrc\n", VINF_SUCCESS));
2065	return VINF_SUCCESS;
2066	}
2067
2068	/** @copydoc VDIMAGEBACKEND::pfnRegionListRelease */
2069	static DECLCALLBACK(void) qedRegionListRelease(void *pBackendData, PCVDREGIONLIST pRegionList)
2070	{
2071	RT_NOREF1(pRegionList);
2072	LogFlowFunc(("pBackendData=%#p pRegionList=%#p\n", pBackendData, pRegionList));
2073	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2074	AssertPtr(pThis); RT_NOREF(pThis);
2075
2076	/* Nothing to do here. */
2077	}
2078
2079	/** @copydoc VDIMAGEBACKEND::pfnGetImageFlags */
2080	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2081	{
2082	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2083	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2084
2085	AssertPtrReturn(pImage, 0);
2086
2087	LogFlowFunc(("returns %#x\n", pImage->uImageFlags));
2088	return pImage->uImageFlags;
2089	}
2090
2091	/** @copydoc VDIMAGEBACKEND::pfnGetOpenFlags */
2092	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2093	{
2094	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2095	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2096
2097	AssertPtrReturn(pImage, 0);
2098
2099	LogFlowFunc(("returns %#x\n", pImage->uOpenFlags));
2100	return pImage->uOpenFlags;
2101	}
2102
2103	/** @copydoc VDIMAGEBACKEND::pfnSetOpenFlags */
2104	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2105	{
2106	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2107	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2108	int rc = VINF_SUCCESS;
2109
2110	/* Image must be opened and the new flags must be valid. */
2111	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2112	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SHAREABLE
2113	\| VD_OPEN_FLAGS_SEQUENTIAL \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2114	rc = VERR_INVALID_PARAMETER;
2115	else
2116	{
2117	/* Implement this operation via reopening the image. */
2118	rc = qedFreeImage(pImage, false);
2119	if (RT_SUCCESS(rc))
2120	rc = qedOpenImage(pImage, uOpenFlags);
2121	}
2122
2123	LogFlowFunc(("returns %Rrc\n", rc));
2124	return rc;
2125	}
2126
2127	/** @copydoc VDIMAGEBACKEND::pfnGetComment */
2128	VD_BACKEND_CALLBACK_GET_COMMENT_DEF_NOT_SUPPORTED(qedGetComment);
2129
2130	/** @copydoc VDIMAGEBACKEND::pfnSetComment */
2131	VD_BACKEND_CALLBACK_SET_COMMENT_DEF_NOT_SUPPORTED(qedSetComment, PQEDIMAGE);
2132
2133	/** @copydoc VDIMAGEBACKEND::pfnGetUuid */
2134	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetUuid);
2135
2136	/** @copydoc VDIMAGEBACKEND::pfnSetUuid */
2137	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetUuid, PQEDIMAGE);
2138
2139	/** @copydoc VDIMAGEBACKEND::pfnGetModificationUuid */
2140	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetModificationUuid);
2141
2142	/** @copydoc VDIMAGEBACKEND::pfnSetModificationUuid */
2143	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetModificationUuid, PQEDIMAGE);
2144
2145	/** @copydoc VDIMAGEBACKEND::pfnGetParentUuid */
2146	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetParentUuid);
2147
2148	/** @copydoc VDIMAGEBACKEND::pfnSetParentUuid */
2149	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetParentUuid, PQEDIMAGE);
2150
2151	/** @copydoc VDIMAGEBACKEND::pfnGetParentModificationUuid */
2152	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetParentModificationUuid);
2153
2154	/** @copydoc VDIMAGEBACKEND::pfnSetParentModificationUuid */
2155	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetParentModificationUuid, PQEDIMAGE);
2156
2157	/** @copydoc VDIMAGEBACKEND::pfnDump */
2158	static DECLCALLBACK(void) qedDump(void *pBackendData)
2159	{
2160	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2161
2162	AssertPtrReturnVoid(pImage);
2163	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2164	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2165	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2166	pImage->cbSize / 512);
2167	}
2168
2169	/** @copydoc VDIMAGEBACKEND::pfnGetParentFilename */
2170	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2171	{
2172	int rc = VINF_SUCCESS;
2173	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2174
2175	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2176
2177	if (pImage->pszBackingFilename)
2178	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2179	else
2180	rc = VERR_NOT_SUPPORTED;
2181
2182	LogFlowFunc(("returns %Rrc\n", rc));
2183	return rc;
2184	}
2185
2186	/** @copydoc VDIMAGEBACKEND::pfnSetParentFilename */
2187	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2188	{
2189	int rc = VINF_SUCCESS;
2190	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2191
2192	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2193
2194	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2195	rc = VERR_VD_IMAGE_READ_ONLY;
2196	else if ( pImage->pszBackingFilename
2197	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2198	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2199	else
2200	{
2201	if (pImage->pszBackingFilename)
2202	RTStrFree(pImage->pszBackingFilename);
2203	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2204	if (!pImage->pszBackingFilename)
2205	rc = VERR_NO_MEMORY;
2206	else
2207	{
2208	if (!pImage->offBackingFilename)
2209	{
2210	/* Allocate new cluster. */
2211	uint64_t offData = qedClusterAllocate(pImage, 1);
2212
2213	Assert((offData & UINT32_MAX) == offData);
2214	pImage->offBackingFilename = (uint32_t)offData;
2215	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2216	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2217	offData + pImage->cbCluster);
2218	}
2219
2220	if (RT_SUCCESS(rc))
2221	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2222	pImage->offBackingFilename,
2223	pImage->pszBackingFilename,
2224	strlen(pImage->pszBackingFilename));
2225	}
2226	}
2227
2228	LogFlowFunc(("returns %Rrc\n", rc));
2229	return rc;
2230	}
2231
2232	/** @copydoc VDIMAGEBACKEND::pfnResize */
2233	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2234	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2235	unsigned uPercentStart, unsigned uPercentSpan,
2236	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2237	PVDINTERFACE pVDIfsOperation)
2238	{
2239	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2240	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2241	int rc = VINF_SUCCESS;
2242
2243	/* Making the image smaller is not supported at the moment. */
2244	if (cbSize < pImage->cbSize)
2245	rc = VERR_NOT_SUPPORTED;
2246	else if (cbSize > pImage->cbSize)
2247	{
2248	/*
2249	* It is enough to just update the size field in the header to complete
2250	* growing. With the default cluster and table sizes the image can be expanded
2251	* to 64TB without overflowing the L1 and L2 tables making block relocation
2252	* superfluous.
2253	* @todo: The rare case where block relocation is still required (non default
2254	* table and/or cluster size or images with more than 64TB) is not
2255	* implemented yet and resizing such an image will fail with an error.
2256	*/
2257	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2258	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2259	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2260	pImage->pszFilename);
2261	else
2262	{
2263	uint64_t cbSizeOld = pImage->cbSize;
2264
2265	pImage->cbSize = cbSize;
2266	rc = qedFlushImage(pImage);
2267	if (RT_FAILURE(rc))
2268	{
2269	pImage->cbSize = cbSizeOld; /* Restore */
2270
2271	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2272	pImage->pszFilename);
2273	}
2274	}
2275	}
2276	/* Same size doesn't change the image at all. */
2277
2278	LogFlowFunc(("returns %Rrc\n", rc));
2279	return rc;
2280	}
2281
2282
2283	const VDIMAGEBACKEND g_QedBackend =
2284	{
2285	/* u32Version */
2286	VD_IMGBACKEND_VERSION,
2287	/* pszBackendName */
2288	"QED",
2289	/* uBackendCaps */
2290	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2291	/* paFileExtensions */
2292	s_aQedFileExtensions,
2293	/* paConfigInfo */
2294	NULL,
2295	/* pfnProbe */
2296	qedProbe,
2297	/* pfnOpen */
2298	qedOpen,
2299	/* pfnCreate */
2300	qedCreate,
2301	/* pfnRename */
2302	qedRename,
2303	/* pfnClose */
2304	qedClose,
2305	/* pfnRead */
2306	qedRead,
2307	/* pfnWrite */
2308	qedWrite,
2309	/* pfnFlush */
2310	qedFlush,
2311	/* pfnDiscard */
2312	NULL,
2313	/* pfnGetVersion */
2314	qedGetVersion,
2315	/* pfnGetFileSize */
2316	qedGetFileSize,
2317	/* pfnGetPCHSGeometry */
2318	qedGetPCHSGeometry,
2319	/* pfnSetPCHSGeometry */
2320	qedSetPCHSGeometry,
2321	/* pfnGetLCHSGeometry */
2322	qedGetLCHSGeometry,
2323	/* pfnSetLCHSGeometry */
2324	qedSetLCHSGeometry,
2325	/* pfnQueryRegions */
2326	qedQueryRegions,
2327	/* pfnRegionListRelease */
2328	qedRegionListRelease,
2329	/* pfnGetImageFlags */
2330	qedGetImageFlags,
2331	/* pfnGetOpenFlags */
2332	qedGetOpenFlags,
2333	/* pfnSetOpenFlags */
2334	qedSetOpenFlags,
2335	/* pfnGetComment */
2336	qedGetComment,
2337	/* pfnSetComment */
2338	qedSetComment,
2339	/* pfnGetUuid */
2340	qedGetUuid,
2341	/* pfnSetUuid */
2342	qedSetUuid,
2343	/* pfnGetModificationUuid */
2344	qedGetModificationUuid,
2345	/* pfnSetModificationUuid */
2346	qedSetModificationUuid,
2347	/* pfnGetParentUuid */
2348	qedGetParentUuid,
2349	/* pfnSetParentUuid */
2350	qedSetParentUuid,
2351	/* pfnGetParentModificationUuid */
2352	qedGetParentModificationUuid,
2353	/* pfnSetParentModificationUuid */
2354	qedSetParentModificationUuid,
2355	/* pfnDump */
2356	qedDump,
2357	/* pfnGetTimestamp */
2358	NULL,
2359	/* pfnGetParentTimestamp */
2360	NULL,
2361	/* pfnSetParentTimestamp */
2362	NULL,
2363	/* pfnGetParentFilename */
2364	qedGetParentFilename,
2365	/* pfnSetParentFilename */
2366	qedSetParentFilename,
2367	/* pfnComposeLocation */
2368	genericFileComposeLocation,
2369	/* pfnComposeName */
2370	genericFileComposeName,
2371	/* pfnCompact */
2372	NULL,
2373	/* pfnResize */
2374	qedResize,
2375	/* pfnRepair */
2376	NULL,
2377	/* pfnTraverseMetadata */
2378	NULL,
2379	/* u32Version */
2380	VD_IMGBACKEND_VERSION
2381	};

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 69500

Download in other formats: