QED.cpp@ 95140

Last change on this file since 95140 was 93115, checked in by vboxsync, 3 years ago
scm --update-copyright-year
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 86.8 KB

Line
1	/* $Id: QED.cpp 93115 2022-01-01 11:31:46Z vboxsync $ */
2	/** @file
3	* QED - QED Disk image.
4	*/
5
6	/*
7	* Copyright (C) 2011-2022 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	RTStrFree(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 = RTStrAlloc(Header.u32BackingFilenameSize + 1); /* +1 for \0 terminator. */
1178	if (pImage->pszBackingFilename)
1179	{
1180	RT_BZERO(pImage->pszBackingFilename, Header.u32BackingFilenameSize + 1);
1181	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1182	pImage->offBackingFilename = Header.u32OffBackingFilename;
1183	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1184	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1185	Header.u32BackingFilenameSize);
1186	if (RT_SUCCESS(rc))
1187	rc = RTStrValidateEncoding(pImage->pszBackingFilename);
1188	}
1189	else
1190	rc = VERR_NO_STR_MEMORY;
1191	}
1192
1193	if (RT_SUCCESS(rc))
1194	{
1195	pImage->cbImage = cbFile;
1196	pImage->cbCluster = Header.u32ClusterSize;
1197	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1198	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1199	pImage->offL1Table = Header.u64OffL1Table;
1200	pImage->cbSize = Header.u64Size;
1201	qedTableMasksInit(pImage);
1202
1203	/* Allocate L1 table. */
1204	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1205	if (pImage->paL1Table)
1206	{
1207	/* Read from the image. */
1208	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1209	pImage->offL1Table, pImage->paL1Table,
1210	pImage->cbTable);
1211	if (RT_SUCCESS(rc))
1212	{
1213	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1214
1215	/* If the consistency check succeeded, clear the flag by flushing the image. */
1216	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1217	rc = qedFlushImage(pImage);
1218	}
1219	else
1220	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1221	N_("Qed: Reading the L1 table for image '%s' failed"),
1222	pImage->pszFilename);
1223	}
1224	else
1225	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1226	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1227	pImage->pszFilename);
1228	}
1229	}
1230	else
1231	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1232	N_("Qed: The image '%s' makes use of unsupported features"),
1233	pImage->pszFilename);
1234	}
1235	else if (RT_SUCCESS(rc))
1236	rc = VERR_VD_GEN_INVALID_HEADER;
1237	}
1238	else if (RT_SUCCESS(rc))
1239	rc = VERR_VD_GEN_INVALID_HEADER;
1240	}
1241	/* else: Do NOT signal an appropriate error here, as the VD layer has the
1242	* choice of retrying the open if it failed. */
1243	}
1244	else
1245	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1246	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1247	pImage->pszFilename);
1248
1249	if (RT_SUCCESS(rc))
1250	{
1251	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1252	pImage->RegionList.fFlags = 0;
1253	pImage->RegionList.cRegions = 1;
1254
1255	pRegion->offRegion = 0; /* Disk start. */
1256	pRegion->cbBlock = 512;
1257	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1258	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1259	pRegion->cbData = 512;
1260	pRegion->cbMetadata = 0;
1261	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1262	}
1263	else
1264	qedFreeImage(pImage, false);
1265	return rc;
1266	}
1267
1268	/**
1269	* Internal: Create a qed image.
1270	*/
1271	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1272	unsigned uImageFlags, const char *pszComment,
1273	PCVDGEOMETRY pPCHSGeometry,
1274	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1275	PVDINTERFACEPROGRESS pIfProgress,
1276	unsigned uPercentStart, unsigned uPercentSpan)
1277	{
1278	RT_NOREF1(pszComment);
1279	int rc;
1280
1281	if (!(uImageFlags & VD_IMAGE_FLAGS_FIXED))
1282	{
1283	rc = qedL2TblCacheCreate(pImage);
1284	if (RT_SUCCESS(rc))
1285	{
1286	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1287	pImage->uImageFlags = uImageFlags;
1288	pImage->PCHSGeometry = *pPCHSGeometry;
1289	pImage->LCHSGeometry = *pLCHSGeometry;
1290
1291	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1292	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1293	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1294
1295	/* Create image file. */
1296	uint32_t fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1297	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1298	if (RT_SUCCESS(rc))
1299	{
1300	/* Init image state. */
1301	pImage->cbSize = cbSize;
1302	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1303	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1304	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1305	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1306	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1307	pImage->cbBackingFilename = 0;
1308	pImage->offBackingFilename = 0;
1309	qedTableMasksInit(pImage);
1310
1311	/* Init L1 table. */
1312	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1313	if (RT_LIKELY(pImage->paL1Table))
1314	{
1315	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan * 98 / 100);
1316	rc = qedFlushImage(pImage);
1317	}
1318	else
1319	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1320	pImage->pszFilename);
1321	}
1322	else
1323	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1324	}
1325	else
1326	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1327	pImage->pszFilename);
1328	}
1329	else
1330	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1331
1332	if (RT_SUCCESS(rc))
1333	{
1334	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1335	pImage->RegionList.fFlags = 0;
1336	pImage->RegionList.cRegions = 1;
1337
1338	pRegion->offRegion = 0; /* Disk start. */
1339	pRegion->cbBlock = 512;
1340	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1341	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1342	pRegion->cbData = 512;
1343	pRegion->cbMetadata = 0;
1344	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1345
1346	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan);
1347	}
1348	else
1349	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1350
1351	return rc;
1352	}
1353
1354	/**
1355	* Rollback anything done during async cluster allocation.
1356	*
1357	* @returns VBox status code.
1358	* @param pImage The image instance data.
1359	* @param pIoCtx The I/O context.
1360	* @param pClusterAlloc The cluster allocation to rollback.
1361	*/
1362	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1363	{
1364	RT_NOREF1(pIoCtx);
1365	int rc = VINF_SUCCESS;
1366
1367	switch (pClusterAlloc->enmAllocState)
1368	{
1369	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1370	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1371	{
1372	/* Revert the L1 table entry */
1373	pImage->paL1Table[pClusterAlloc->idxL1] = 0;
1374	pImage->pL2TblAlloc = NULL;
1375
1376	/* Assumption right now is that the L1 table is not modified on storage if the link fails. */
1377	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1378	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1379	Assert(!pClusterAlloc->pL2Entry->cRefs);
1380	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1381	break;
1382	}
1383	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1384	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1385	{
1386	/* Assumption right now is that the L2 table is not modified if the link fails. */
1387	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = 0;
1388	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1389	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1390	break;
1391	}
1392	default:
1393	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1394	rc = VERR_INVALID_STATE;
1395	}
1396
1397	RTMemFree(pClusterAlloc);
1398	return rc;
1399	}
1400
1401	/**
1402	* Updates the state of the async cluster allocation.
1403	*
1404	* @returns VBox status code.
1405	* @param pBackendData The opaque backend data.
1406	* @param pIoCtx I/O context associated with this request.
1407	* @param pvUser Opaque user data passed during a read/write request.
1408	* @param rcReq Status code for the completed request.
1409	*/
1410	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1411	{
1412	int rc = VINF_SUCCESS;
1413	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1414	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1415
1416	if (RT_FAILURE(rcReq))
1417	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1418
1419	AssertPtr(pClusterAlloc->pL2Entry);
1420
1421	switch (pClusterAlloc->enmAllocState)
1422	{
1423	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1424	{
1425	/* Update the link in the in memory L1 table now. */
1426	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1427
1428	/* Update the link in the on disk L1 table now. */
1429	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1430	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1431	qedAsyncClusterAllocUpdate, pClusterAlloc);
1432	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1433	break;
1434	else if (RT_FAILURE(rc))
1435	{
1436	/* Rollback. */
1437	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1438	break;
1439	}
1440	}
1441	RT_FALL_THRU();
1442	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1443	{
1444	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1445	uint64_t offData = qedClusterAllocate(pImage, 1);
1446
1447	pImage->pL2TblAlloc = NULL;
1448	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1449
1450	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1451	pClusterAlloc->cbImageOld = offData;
1452	pClusterAlloc->offClusterNew = offData;
1453
1454	/* Write data. */
1455	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1456	offData, pIoCtx, pClusterAlloc->cbToWrite,
1457	qedAsyncClusterAllocUpdate, pClusterAlloc);
1458	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1459	break;
1460	else if (RT_FAILURE(rc))
1461	{
1462	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1463	RTMemFree(pClusterAlloc);
1464	break;
1465	}
1466	}
1467	RT_FALL_THRU();
1468	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1469	{
1470	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1471	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1472
1473	/* Link L2 table and update it. */
1474	rc = qedTblWrite(pImage, pIoCtx, pImage->paL1Table[pClusterAlloc->idxL1],
1475	pClusterAlloc->pL2Entry->paL2Tbl,
1476	qedAsyncClusterAllocUpdate, pClusterAlloc);
1477	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1478	break;
1479	else if (RT_FAILURE(rc))
1480	{
1481	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1482	RTMemFree(pClusterAlloc);
1483	break;
1484	}
1485	}
1486	RT_FALL_THRU();
1487	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1488	{
1489	/* Everything done without errors, signal completion. */
1490	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1491	RTMemFree(pClusterAlloc);
1492	rc = VINF_SUCCESS;
1493	break;
1494	}
1495	default:
1496	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1497	pClusterAlloc->enmAllocState));
1498	}
1499
1500	return rc;
1501	}
1502
1503	/** @copydoc VDIMAGEBACKEND::pfnProbe */
1504	static DECLCALLBACK(int) qedProbe(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1505	PVDINTERFACE pVDIfsImage, VDTYPE enmDesiredType, VDTYPE *penmType)
1506	{
1507	RT_NOREF(pVDIfsDisk, enmDesiredType);
1508	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1509	PVDIOSTORAGE pStorage = NULL;
1510	int rc = VINF_SUCCESS;
1511
1512	/* Get I/O interface. */
1513	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1514	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1515	AssertPtrReturn(pszFilename, VERR_INVALID_POINTER);
1516	AssertReturn(*pszFilename != '\0', VERR_INVALID_PARAMETER);
1517
1518
1519	/*
1520	* Open the file and read the footer.
1521	*/
1522	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1523	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1524	false /* fCreate */),
1525	&pStorage);
1526	if (RT_SUCCESS(rc))
1527	{
1528	uint64_t cbFile;
1529
1530	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1531	if ( RT_SUCCESS(rc)
1532	&& cbFile > sizeof(QedHeader))
1533	{
1534	QedHeader Header;
1535
1536	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1537	if ( RT_SUCCESS(rc)
1538	&& qedHdrConvertToHostEndianess(&Header))
1539	*penmType = VDTYPE_HDD;
1540	else
1541	rc = VERR_VD_GEN_INVALID_HEADER;
1542	}
1543	else
1544	rc = VERR_VD_GEN_INVALID_HEADER;
1545	}
1546
1547	if (pStorage)
1548	vdIfIoIntFileClose(pIfIo, pStorage);
1549
1550	LogFlowFunc(("returns %Rrc\n", rc));
1551	return rc;
1552	}
1553
1554	/** @copydoc VDIMAGEBACKEND::pfnOpen */
1555	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1556	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1557	VDTYPE enmType, void **ppBackendData)
1558	{
1559	RT_NOREF1(enmType); /*< @todo r=klaus make use of the type info. /
1560
1561	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n",
1562	pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1563	int rc;
1564
1565	/* Check open flags. All valid flags are supported. */
1566	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1567	AssertPtrReturn(pszFilename, VERR_INVALID_POINTER);
1568	AssertReturn(*pszFilename != '\0', VERR_INVALID_PARAMETER);
1569
1570
1571	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1572	if (RT_LIKELY(pImage))
1573	{
1574	pImage->pszFilename = pszFilename;
1575	pImage->pStorage = NULL;
1576	pImage->pVDIfsDisk = pVDIfsDisk;
1577	pImage->pVDIfsImage = pVDIfsImage;
1578
1579	rc = qedOpenImage(pImage, uOpenFlags);
1580	if (RT_SUCCESS(rc))
1581	*ppBackendData = pImage;
1582	else
1583	RTMemFree(pImage);
1584	}
1585	else
1586	rc = VERR_NO_MEMORY;
1587
1588	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1589	return rc;
1590	}
1591
1592	/** @copydoc VDIMAGEBACKEND::pfnCreate */
1593	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1594	unsigned uImageFlags, const char *pszComment,
1595	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1596	PCRTUUID pUuid, unsigned uOpenFlags,
1597	unsigned uPercentStart, unsigned uPercentSpan,
1598	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1599	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1600	void **ppBackendData)
1601	{
1602	RT_NOREF1(pUuid);
1603	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",
1604	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1605	int rc;
1606
1607	/* Check the VD container type. */
1608	if (enmType != VDTYPE_HDD)
1609	return VERR_VD_INVALID_TYPE;
1610
1611	/* Check open flags. All valid flags are supported. */
1612	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1613	AssertPtrReturn(pszFilename, VERR_INVALID_POINTER);
1614	AssertReturn(*pszFilename != '\0', VERR_INVALID_PARAMETER);
1615	AssertPtrReturn(pPCHSGeometry, VERR_INVALID_POINTER);
1616	AssertPtrReturn(pLCHSGeometry, VERR_INVALID_POINTER);
1617
1618	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1619	if (RT_LIKELY(pImage))
1620	{
1621	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1622
1623	pImage->pszFilename = pszFilename;
1624	pImage->pStorage = NULL;
1625	pImage->pVDIfsDisk = pVDIfsDisk;
1626	pImage->pVDIfsImage = pVDIfsImage;
1627
1628	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1629	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1630	pIfProgress, uPercentStart, uPercentSpan);
1631	if (RT_SUCCESS(rc))
1632	{
1633	/* So far the image is opened in read/write mode. Make sure the
1634	* image is opened in read-only mode if the caller requested that. */
1635	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1636	{
1637	qedFreeImage(pImage, false);
1638	rc = qedOpenImage(pImage, uOpenFlags);
1639	}
1640
1641	if (RT_SUCCESS(rc))
1642	*ppBackendData = pImage;
1643	}
1644
1645	if (RT_FAILURE(rc))
1646	RTMemFree(pImage);
1647	}
1648	else
1649	rc = VERR_NO_MEMORY;
1650
1651	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1652	return rc;
1653	}
1654
1655	/** @copydoc VDIMAGEBACKEND::pfnRename */
1656	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1657	{
1658	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1659	int rc = VINF_SUCCESS;
1660	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1661
1662	/* Check arguments. */
1663	AssertReturn((pImage && pszFilename && *pszFilename), VERR_INVALID_PARAMETER);
1664
1665	/* Close the image. */
1666	rc = qedFreeImage(pImage, false);
1667	if (RT_SUCCESS(rc))
1668	{
1669	/* Rename the file. */
1670	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1671	if (RT_SUCCESS(rc))
1672	{
1673	/* Update pImage with the new information. */
1674	pImage->pszFilename = pszFilename;
1675
1676	/* Open the old image with new name. */
1677	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1678	}
1679	else
1680	{
1681	/* The move failed, try to reopen the original image. */
1682	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1683	if (RT_FAILURE(rc2))
1684	rc = rc2;
1685	}
1686	}
1687
1688	LogFlowFunc(("returns %Rrc\n", rc));
1689	return rc;
1690	}
1691
1692	/** @copydoc VDIMAGEBACKEND::pfnClose */
1693	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1694	{
1695	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1696	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1697
1698	int rc = qedFreeImage(pImage, fDelete);
1699	RTMemFree(pImage);
1700
1701	LogFlowFunc(("returns %Rrc\n", rc));
1702	return rc;
1703	}
1704
1705	/** @copydoc VDIMAGEBACKEND::pfnRead */
1706	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1707	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1708	{
1709	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1710	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1711	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1712	uint32_t offCluster = 0;
1713	uint32_t idxL1 = 0;
1714	uint32_t idxL2 = 0;
1715	uint64_t offFile = 0;
1716
1717	AssertPtr(pImage);
1718	Assert(uOffset % 512 == 0);
1719	Assert(cbToRead % 512 == 0);
1720	AssertPtrReturn(pIoCtx, VERR_INVALID_POINTER);
1721	AssertReturn(cbToRead, VERR_INVALID_PARAMETER);
1722	AssertReturn(uOffset + cbToRead <= pImage->cbSize, VERR_INVALID_PARAMETER);
1723
1724	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1725
1726	/* Clip read size to remain in the cluster. */
1727	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1728
1729	/* Get offset in image. */
1730	int rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1731	if (RT_SUCCESS(rc))
1732	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1733	pIoCtx, cbToRead);
1734
1735	if ( ( RT_SUCCESS(rc)
1736	\|\| rc == VERR_VD_BLOCK_FREE
1737	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1738	&& pcbActuallyRead)
1739	*pcbActuallyRead = cbToRead;
1740
1741	LogFlowFunc(("returns %Rrc\n", rc));
1742	return rc;
1743	}
1744
1745	/** @copydoc VDIMAGEBACKEND::pfnWrite */
1746	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1747	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1748	size_t *pcbPostRead, unsigned fWrite)
1749	{
1750	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1751	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1752	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1753	uint32_t offCluster = 0;
1754	uint32_t idxL1 = 0;
1755	uint32_t idxL2 = 0;
1756	uint64_t offImage = 0;
1757	int rc = VINF_SUCCESS;
1758
1759	AssertPtr(pImage);
1760	Assert(!(uOffset % 512));
1761	Assert(!(cbToWrite % 512));
1762	AssertPtrReturn(pIoCtx, VERR_INVALID_POINTER);
1763	AssertReturn(cbToWrite, VERR_INVALID_PARAMETER);
1764	AssertReturn(uOffset + cbToWrite <= pImage->cbSize, VERR_INVALID_PARAMETER);
1765
1766	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1767	{
1768	/* Convert offset to L1, L2 index and cluster offset. */
1769	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1770
1771	/* Clip write size to remain in the cluster. */
1772	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1773	Assert(!(cbToWrite % 512));
1774
1775	/* Get offset in image. */
1776	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1777	if (RT_SUCCESS(rc))
1778	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1779	offImage, pIoCtx, cbToWrite, NULL, NULL);
1780	else if (rc == VERR_VD_BLOCK_FREE)
1781	{
1782	if ( cbToWrite == pImage->cbCluster
1783	&& !(fWrite & VD_WRITE_NO_ALLOC))
1784	{
1785	PQEDL2CACHEENTRY pL2Entry = NULL;
1786
1787	/* Full cluster write to previously unallocated cluster.
1788	* Allocate cluster and write data. */
1789	Assert(!offCluster);
1790
1791	do
1792	{
1793	/* Check if we have to allocate a new cluster for L2 tables. */
1794	if (!pImage->paL1Table[idxL1])
1795	{
1796	uint64_t offL2Tbl;
1797	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1798
1799	/* Allocate new async cluster allocation state. */
1800	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1801	if (RT_UNLIKELY(!pL2ClusterAlloc))
1802	{
1803	rc = VERR_NO_MEMORY;
1804	break;
1805	}
1806
1807	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1808	if (!pL2Entry)
1809	{
1810	rc = VERR_NO_MEMORY;
1811	RTMemFree(pL2ClusterAlloc);
1812	break;
1813	}
1814
1815	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1816	pL2Entry->offL2Tbl = offL2Tbl;
1817	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1818
1819	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1820	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1821	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1822	pL2ClusterAlloc->idxL1 = idxL1;
1823	pL2ClusterAlloc->idxL2 = idxL2;
1824	pL2ClusterAlloc->cbToWrite = cbToWrite;
1825	pL2ClusterAlloc->pL2Entry = pL2Entry;
1826
1827	pImage->pL2TblAlloc = pL2Entry;
1828
1829	LogFlowFunc(("Allocating new L2 table at cluster offset %llu\n", offL2Tbl));
1830
1831	/*
1832	* Write the L2 table first and link to the L1 table afterwards.
1833	* If something unexpected happens the worst case which can happen
1834	* is a leak of some clusters.
1835	*/
1836	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1837	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1838	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1839	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1840	break;
1841	else if (RT_FAILURE(rc))
1842	{
1843	RTMemFree(pL2ClusterAlloc);
1844	qedL2TblCacheEntryFree(pImage, pL2Entry);
1845	break;
1846	}
1847
1848	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1849	}
1850	else
1851	{
1852	LogFlowFunc(("Fetching L2 table at cluster offset %llu\n", pImage->paL1Table[idxL1]));
1853
1854	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1855	&pL2Entry);
1856
1857	if (RT_SUCCESS(rc))
1858	{
1859	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1860
1861	/* Allocate new async cluster allocation state. */
1862	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1863	if (RT_UNLIKELY(!pDataClusterAlloc))
1864	{
1865	rc = VERR_NO_MEMORY;
1866	break;
1867	}
1868
1869	/* Allocate new cluster for the data. */
1870	uint64_t offData = qedClusterAllocate(pImage, 1);
1871
1872	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1873	pDataClusterAlloc->cbImageOld = offData;
1874	pDataClusterAlloc->offClusterNew = offData;
1875	pDataClusterAlloc->idxL1 = idxL1;
1876	pDataClusterAlloc->idxL2 = idxL2;
1877	pDataClusterAlloc->cbToWrite = cbToWrite;
1878	pDataClusterAlloc->pL2Entry = pL2Entry;
1879
1880	/* Write data. */
1881	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1882	offData, pIoCtx, cbToWrite,
1883	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1884	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1885	break;
1886	else if (RT_FAILURE(rc))
1887	{
1888	RTMemFree(pDataClusterAlloc);
1889	break;
1890	}
1891
1892	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1893	}
1894	}
1895
1896	} while (0);
1897
1898	*pcbPreRead = 0;
1899	*pcbPostRead = 0;
1900	}
1901	else
1902	{
1903	/* Trying to do a partial write to an unallocated cluster. Don't do
1904	* anything except letting the upper layer know what to do. */
1905	*pcbPreRead = offCluster;
1906	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
1907	}
1908	}
1909
1910	if (pcbWriteProcess)
1911	*pcbWriteProcess = cbToWrite;
1912	}
1913	else
1914	rc = VERR_VD_IMAGE_READ_ONLY;
1915
1916	LogFlowFunc(("returns %Rrc\n", rc));
1917	return rc;
1918	}
1919
1920	/** @copydoc VDIMAGEBACKEND::pfnFlush */
1921	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
1922	{
1923	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1924	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1925	int rc = VINF_SUCCESS;
1926
1927	AssertPtr(pImage);
1928	AssertPtrReturn(pIoCtx, VERR_INVALID_PARAMETER);
1929
1930	if ( pImage->pStorage
1931	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1932	{
1933	QedHeader Header;
1934
1935	Assert(!(pImage->cbTable % pImage->cbCluster));
1936	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1937	NULL, NULL);
1938	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1939	{
1940	/* Write header. */
1941	qedHdrConvertFromHostEndianess(pImage, &Header);
1942	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1943	0, &Header, sizeof(Header),
1944	pIoCtx, NULL, NULL);
1945	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1946	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
1947	pIoCtx, NULL, NULL);
1948	}
1949	}
1950
1951	LogFlowFunc(("returns %Rrc\n", rc));
1952	return rc;
1953	}
1954
1955	/** @copydoc VDIMAGEBACKEND::pfnGetVersion */
1956	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
1957	{
1958	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1959	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1960
1961	AssertPtrReturn(pImage, 0);
1962
1963	return 1;
1964	}
1965
1966	/** @copydoc VDIMAGEBACKEND::pfnGetFileSize */
1967	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
1968	{
1969	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1970	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1971	uint64_t cb = 0;
1972
1973	AssertPtrReturn(pImage, 0);
1974
1975	uint64_t cbFile;
1976	if (pImage->pStorage)
1977	{
1978	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1979	if (RT_SUCCESS(rc))
1980	cb += cbFile;
1981	}
1982
1983	LogFlowFunc(("returns %lld\n", cb));
1984	return cb;
1985	}
1986
1987	/** @copydoc VDIMAGEBACKEND::pfnGetPCHSGeometry */
1988	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
1989	PVDGEOMETRY pPCHSGeometry)
1990	{
1991	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
1992	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1993	int rc = VINF_SUCCESS;
1994
1995	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
1996
1997	if (pImage->PCHSGeometry.cCylinders)
1998	*pPCHSGeometry = pImage->PCHSGeometry;
1999	else
2000	rc = VERR_VD_GEOMETRY_NOT_SET;
2001
2002	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2003	return rc;
2004	}
2005
2006	/** @copydoc VDIMAGEBACKEND::pfnSetPCHSGeometry */
2007	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
2008	PCVDGEOMETRY pPCHSGeometry)
2009	{
2010	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n",
2011	pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2012	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2013	int rc = VINF_SUCCESS;
2014
2015	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2016
2017	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2018	rc = VERR_VD_IMAGE_READ_ONLY;
2019	else
2020	pImage->PCHSGeometry = *pPCHSGeometry;
2021
2022	LogFlowFunc(("returns %Rrc\n", rc));
2023	return rc;
2024	}
2025
2026	/** @copydoc VDIMAGEBACKEND::pfnGetLCHSGeometry */
2027	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData, PVDGEOMETRY pLCHSGeometry)
2028	{
2029	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2030	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2031	int rc = VINF_SUCCESS;
2032
2033	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2034
2035	if (pImage->LCHSGeometry.cCylinders)
2036	*pLCHSGeometry = pImage->LCHSGeometry;
2037	else
2038	rc = VERR_VD_GEOMETRY_NOT_SET;
2039
2040	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders,
2041	pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2042	return rc;
2043	}
2044
2045	/** @copydoc VDIMAGEBACKEND::pfnSetLCHSGeometry */
2046	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData, PCVDGEOMETRY pLCHSGeometry)
2047	{
2048	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData,
2049	pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2050	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2051	int rc = VINF_SUCCESS;
2052
2053	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2054
2055	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2056	rc = VERR_VD_IMAGE_READ_ONLY;
2057	else
2058	pImage->LCHSGeometry = *pLCHSGeometry;
2059
2060	LogFlowFunc(("returns %Rrc\n", rc));
2061	return rc;
2062	}
2063
2064	/** @copydoc VDIMAGEBACKEND::pfnQueryRegions */
2065	static DECLCALLBACK(int) qedQueryRegions(void pBackendData, PCVDREGIONLIST ppRegionList)
2066	{
2067	LogFlowFunc(("pBackendData=%#p ppRegionList=%#p\n", pBackendData, ppRegionList));
2068	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2069
2070	AssertPtrReturn(pThis, VERR_VD_NOT_OPENED);
2071
2072	*ppRegionList = &pThis->RegionList;
2073	LogFlowFunc(("returns %Rrc\n", VINF_SUCCESS));
2074	return VINF_SUCCESS;
2075	}
2076
2077	/** @copydoc VDIMAGEBACKEND::pfnRegionListRelease */
2078	static DECLCALLBACK(void) qedRegionListRelease(void *pBackendData, PCVDREGIONLIST pRegionList)
2079	{
2080	RT_NOREF1(pRegionList);
2081	LogFlowFunc(("pBackendData=%#p pRegionList=%#p\n", pBackendData, pRegionList));
2082	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2083	AssertPtr(pThis); RT_NOREF(pThis);
2084
2085	/* Nothing to do here. */
2086	}
2087
2088	/** @copydoc VDIMAGEBACKEND::pfnGetImageFlags */
2089	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2090	{
2091	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2092	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2093
2094	AssertPtrReturn(pImage, 0);
2095
2096	LogFlowFunc(("returns %#x\n", pImage->uImageFlags));
2097	return pImage->uImageFlags;
2098	}
2099
2100	/** @copydoc VDIMAGEBACKEND::pfnGetOpenFlags */
2101	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2102	{
2103	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2104	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2105
2106	AssertPtrReturn(pImage, 0);
2107
2108	LogFlowFunc(("returns %#x\n", pImage->uOpenFlags));
2109	return pImage->uOpenFlags;
2110	}
2111
2112	/** @copydoc VDIMAGEBACKEND::pfnSetOpenFlags */
2113	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2114	{
2115	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2116	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2117	int rc = VINF_SUCCESS;
2118
2119	/* Image must be opened and the new flags must be valid. */
2120	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2121	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SHAREABLE
2122	\| VD_OPEN_FLAGS_SEQUENTIAL \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2123	rc = VERR_INVALID_PARAMETER;
2124	else
2125	{
2126	/* Implement this operation via reopening the image. */
2127	rc = qedFreeImage(pImage, false);
2128	if (RT_SUCCESS(rc))
2129	rc = qedOpenImage(pImage, uOpenFlags);
2130	}
2131
2132	LogFlowFunc(("returns %Rrc\n", rc));
2133	return rc;
2134	}
2135
2136	/** @copydoc VDIMAGEBACKEND::pfnGetComment */
2137	VD_BACKEND_CALLBACK_GET_COMMENT_DEF_NOT_SUPPORTED(qedGetComment);
2138
2139	/** @copydoc VDIMAGEBACKEND::pfnSetComment */
2140	VD_BACKEND_CALLBACK_SET_COMMENT_DEF_NOT_SUPPORTED(qedSetComment, PQEDIMAGE);
2141
2142	/** @copydoc VDIMAGEBACKEND::pfnGetUuid */
2143	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetUuid);
2144
2145	/** @copydoc VDIMAGEBACKEND::pfnSetUuid */
2146	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetUuid, PQEDIMAGE);
2147
2148	/** @copydoc VDIMAGEBACKEND::pfnGetModificationUuid */
2149	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetModificationUuid);
2150
2151	/** @copydoc VDIMAGEBACKEND::pfnSetModificationUuid */
2152	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetModificationUuid, PQEDIMAGE);
2153
2154	/** @copydoc VDIMAGEBACKEND::pfnGetParentUuid */
2155	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetParentUuid);
2156
2157	/** @copydoc VDIMAGEBACKEND::pfnSetParentUuid */
2158	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetParentUuid, PQEDIMAGE);
2159
2160	/** @copydoc VDIMAGEBACKEND::pfnGetParentModificationUuid */
2161	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetParentModificationUuid);
2162
2163	/** @copydoc VDIMAGEBACKEND::pfnSetParentModificationUuid */
2164	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetParentModificationUuid, PQEDIMAGE);
2165
2166	/** @copydoc VDIMAGEBACKEND::pfnDump */
2167	static DECLCALLBACK(void) qedDump(void *pBackendData)
2168	{
2169	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2170
2171	AssertPtrReturnVoid(pImage);
2172	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2173	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2174	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2175	pImage->cbSize / 512);
2176	}
2177
2178	/** @copydoc VDIMAGEBACKEND::pfnGetParentFilename */
2179	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2180	{
2181	int rc = VINF_SUCCESS;
2182	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2183
2184	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2185
2186	if (pImage->pszBackingFilename)
2187	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2188	else
2189	rc = VERR_NOT_SUPPORTED;
2190
2191	LogFlowFunc(("returns %Rrc\n", rc));
2192	return rc;
2193	}
2194
2195	/** @copydoc VDIMAGEBACKEND::pfnSetParentFilename */
2196	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2197	{
2198	int rc = VINF_SUCCESS;
2199	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2200
2201	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2202
2203	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2204	rc = VERR_VD_IMAGE_READ_ONLY;
2205	else if ( pImage->pszBackingFilename
2206	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2207	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2208	else
2209	{
2210	if (pImage->pszBackingFilename)
2211	RTStrFree(pImage->pszBackingFilename);
2212	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2213	if (!pImage->pszBackingFilename)
2214	rc = VERR_NO_STR_MEMORY;
2215	else
2216	{
2217	if (!pImage->offBackingFilename)
2218	{
2219	/* Allocate new cluster. */
2220	uint64_t offData = qedClusterAllocate(pImage, 1);
2221
2222	Assert((offData & UINT32_MAX) == offData);
2223	pImage->offBackingFilename = (uint32_t)offData;
2224	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2225	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2226	offData + pImage->cbCluster);
2227	}
2228
2229	if (RT_SUCCESS(rc))
2230	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2231	pImage->offBackingFilename,
2232	pImage->pszBackingFilename,
2233	strlen(pImage->pszBackingFilename));
2234	}
2235	}
2236
2237	LogFlowFunc(("returns %Rrc\n", rc));
2238	return rc;
2239	}
2240
2241	/** @copydoc VDIMAGEBACKEND::pfnResize */
2242	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2243	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2244	unsigned uPercentStart, unsigned uPercentSpan,
2245	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2246	PVDINTERFACE pVDIfsOperation)
2247	{
2248	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2249	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2250	int rc = VINF_SUCCESS;
2251
2252	/* Making the image smaller is not supported at the moment. */
2253	if (cbSize < pImage->cbSize)
2254	rc = VERR_NOT_SUPPORTED;
2255	else if (cbSize > pImage->cbSize)
2256	{
2257	/*
2258	* It is enough to just update the size field in the header to complete
2259	* growing. With the default cluster and table sizes the image can be expanded
2260	* to 64TB without overflowing the L1 and L2 tables making block relocation
2261	* superfluous.
2262	* @todo: The rare case where block relocation is still required (non default
2263	* table and/or cluster size or images with more than 64TB) is not
2264	* implemented yet and resizing such an image will fail with an error.
2265	*/
2266	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2267	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2268	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2269	pImage->pszFilename);
2270	else
2271	{
2272	uint64_t cbSizeOld = pImage->cbSize;
2273
2274	pImage->cbSize = cbSize;
2275	rc = qedFlushImage(pImage);
2276	if (RT_FAILURE(rc))
2277	{
2278	pImage->cbSize = cbSizeOld; /* Restore */
2279
2280	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2281	pImage->pszFilename);
2282	}
2283	}
2284	}
2285	/* Same size doesn't change the image at all. */
2286
2287	LogFlowFunc(("returns %Rrc\n", rc));
2288	return rc;
2289	}
2290
2291
2292	const VDIMAGEBACKEND g_QedBackend =
2293	{
2294	/* u32Version */
2295	VD_IMGBACKEND_VERSION,
2296	/* pszBackendName */
2297	"QED",
2298	/* uBackendCaps */
2299	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2300	/* paFileExtensions */
2301	s_aQedFileExtensions,
2302	/* paConfigInfo */
2303	NULL,
2304	/* pfnProbe */
2305	qedProbe,
2306	/* pfnOpen */
2307	qedOpen,
2308	/* pfnCreate */
2309	qedCreate,
2310	/* pfnRename */
2311	qedRename,
2312	/* pfnClose */
2313	qedClose,
2314	/* pfnRead */
2315	qedRead,
2316	/* pfnWrite */
2317	qedWrite,
2318	/* pfnFlush */
2319	qedFlush,
2320	/* pfnDiscard */
2321	NULL,
2322	/* pfnGetVersion */
2323	qedGetVersion,
2324	/* pfnGetFileSize */
2325	qedGetFileSize,
2326	/* pfnGetPCHSGeometry */
2327	qedGetPCHSGeometry,
2328	/* pfnSetPCHSGeometry */
2329	qedSetPCHSGeometry,
2330	/* pfnGetLCHSGeometry */
2331	qedGetLCHSGeometry,
2332	/* pfnSetLCHSGeometry */
2333	qedSetLCHSGeometry,
2334	/* pfnQueryRegions */
2335	qedQueryRegions,
2336	/* pfnRegionListRelease */
2337	qedRegionListRelease,
2338	/* pfnGetImageFlags */
2339	qedGetImageFlags,
2340	/* pfnGetOpenFlags */
2341	qedGetOpenFlags,
2342	/* pfnSetOpenFlags */
2343	qedSetOpenFlags,
2344	/* pfnGetComment */
2345	qedGetComment,
2346	/* pfnSetComment */
2347	qedSetComment,
2348	/* pfnGetUuid */
2349	qedGetUuid,
2350	/* pfnSetUuid */
2351	qedSetUuid,
2352	/* pfnGetModificationUuid */
2353	qedGetModificationUuid,
2354	/* pfnSetModificationUuid */
2355	qedSetModificationUuid,
2356	/* pfnGetParentUuid */
2357	qedGetParentUuid,
2358	/* pfnSetParentUuid */
2359	qedSetParentUuid,
2360	/* pfnGetParentModificationUuid */
2361	qedGetParentModificationUuid,
2362	/* pfnSetParentModificationUuid */
2363	qedSetParentModificationUuid,
2364	/* pfnDump */
2365	qedDump,
2366	/* pfnGetTimestamp */
2367	NULL,
2368	/* pfnGetParentTimestamp */
2369	NULL,
2370	/* pfnSetParentTimestamp */
2371	NULL,
2372	/* pfnGetParentFilename */
2373	qedGetParentFilename,
2374	/* pfnSetParentFilename */
2375	qedSetParentFilename,
2376	/* pfnComposeLocation */
2377	genericFileComposeLocation,
2378	/* pfnComposeName */
2379	genericFileComposeName,
2380	/* pfnCompact */
2381	NULL,
2382	/* pfnResize */
2383	qedResize,
2384	/* pfnRepair */
2385	NULL,
2386	/* pfnTraverseMetadata */
2387	NULL,
2388	/* u32Version */
2389	VD_IMGBACKEND_VERSION
2390	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 95140

Download in other formats: