QED.cpp@ 63768

Last change on this file since 63768 was 63741, checked in by vboxsync, 8 years ago
Storage/QED,QCOW: Always write the complete L1 and L2 table when allocating a new cluster instead of just the updated entry
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 89.0 KB

Line
1	/* $Id: QED.cpp 63741 2016-09-07 07:53:09Z vboxsync $ */
2	/** @file
3	* QED - QED Disk image.
4	*/
5
6	/*
7	* Copyright (C) 2011-2016 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
36	/**
37	* The QED backend implements support for the qemu enhanced disk format (short QED)
38	* The specification for the format is available under http://wiki.qemu.org/Features/QED/Specification
39	*
40	* Missing things to implement:
41	* - compaction
42	* - resizing which requires block relocation (very rare case)
43	*/
44
45
46	/*********************************************************************************************************************************
47	* Structures in a QED image, little endian *
48	*********************************************************************************************************************************/
49
50	#pragma pack(1)
51	typedef struct QedHeader
52	{
53	/** Magic value. */
54	uint32_t u32Magic;
55	/** Cluster size in bytes. */
56	uint32_t u32ClusterSize;
57	/** Size of L1 and L2 tables in clusters. */
58	uint32_t u32TableSize;
59	/** size of this header structure in clusters. */
60	uint32_t u32HeaderSize;
61	/** Features used for the image. */
62	uint64_t u64FeatureFlags;
63	/** Compatibility features used for the image. */
64	uint64_t u64CompatFeatureFlags;
65	/** Self resetting feature bits. */
66	uint64_t u64AutoresetFeatureFlags;
67	/** Offset of the L1 table in bytes. */
68	uint64_t u64OffL1Table;
69	/** Logical image size as seen by the guest. */
70	uint64_t u64Size;
71	/** Offset of the backing filename in bytes. */
72	uint32_t u32OffBackingFilename;
73	/** Size of the backing filename. */
74	uint32_t u32BackingFilenameSize;
75	} QedHeader;
76	#pragma pack()
77	/** Pointer to a on disk QED header. */
78	typedef QedHeader *PQedHeader;
79
80	/** QED magic value. */
81	#define QED_MAGIC UINT32_C(0x00444551) /* QED\0 */
82	/** Cluster size minimum. */
83	#define QED_CLUSTER_SIZE_MIN RT_BIT(12)
84	/** Cluster size maximum. */
85	#define QED_CLUSTER_SIZE_MAX RT_BIT(26)
86	/** L1 and L2 Table size minimum. */
87	#define QED_TABLE_SIZE_MIN 1
88	/** L1 and L2 Table size maximum. */
89	#define QED_TABLE_SIZE_MAX 16
90
91	/** QED default cluster size when creating an image. */
92	#define QED_CLUSTER_SIZE_DEFAULT (64 * _1K)
93	/** The default table size in clusters. */
94	#define QED_TABLE_SIZE_DEFAULT 4
95
96	/** Feature flags.
97	* @{
98	*/
99	/** Image uses a backing file to provide data for unallocated clusters. */
100	#define QED_FEATURE_BACKING_FILE RT_BIT_64(0)
101	/** Image needs checking before use. */
102	#define QED_FEATURE_NEED_CHECK RT_BIT_64(1)
103	/** Don't probe for format of the backing file, treat as raw image. */
104	#define QED_FEATURE_BACKING_FILE_NO_PROBE RT_BIT_64(2)
105	/** Mask of valid features. */
106	#define QED_FEATURE_MASK (QED_FEATURE_BACKING_FILE \| QED_FEATURE_NEED_CHECK \| QED_FEATURE_BACKING_FILE_NO_PROBE)
107	/** @} */
108
109	/** Compatibility feature flags.
110	* @{
111	*/
112	/** Mask of valid compatibility features. */
113	#define QED_COMPAT_FEATURE_MASK (0)
114	/** @} */
115
116	/** Autoreset feature flags.
117	* @{
118	*/
119	/** Mask of valid autoreset features. */
120	#define QED_AUTORESET_FEATURE_MASK (0)
121	/** @} */
122
123
124	/*********************************************************************************************************************************
125	* Constants And Macros, Structures and Typedefs *
126	*********************************************************************************************************************************/
127
128	/**
129	* QED L2 cache entry.
130	*/
131	typedef struct QEDL2CACHEENTRY
132	{
133	/** List node for the search list. */
134	RTLISTNODE NodeSearch;
135	/** List node for the LRU list. */
136	RTLISTNODE NodeLru;
137	/** Reference counter. */
138	uint32_t cRefs;
139	/** The offset of the L2 table, used as search key. */
140	uint64_t offL2Tbl;
141	/** Pointer to the cached L2 table. */
142	uint64_t *paL2Tbl;
143	} QEDL2CACHEENTRY, *PQEDL2CACHEENTRY;
144
145	/** Maximum amount of memory the cache is allowed to use. */
146	#define QED_L2_CACHE_MEMORY_MAX (2*_1M)
147
148	/**
149	* QED image data structure.
150	*/
151	typedef struct QEDIMAGE
152	{
153	/** Image name. */
154	const char *pszFilename;
155	/** Storage handle. */
156	PVDIOSTORAGE pStorage;
157
158	/** Pointer to the per-disk VD interface list. */
159	PVDINTERFACE pVDIfsDisk;
160	/** Pointer to the per-image VD interface list. */
161	PVDINTERFACE pVDIfsImage;
162	/** Error interface. */
163	PVDINTERFACEERROR pIfError;
164	/** I/O interface. */
165	PVDINTERFACEIOINT pIfIo;
166
167	/** Open flags passed by VBoxHD layer. */
168	unsigned uOpenFlags;
169	/** Image flags defined during creation or determined during open. */
170	unsigned uImageFlags;
171	/** Total size of the image. */
172	uint64_t cbSize;
173	/** Physical geometry of this image. */
174	VDGEOMETRY PCHSGeometry;
175	/** Logical geometry of this image. */
176	VDGEOMETRY LCHSGeometry;
177
178	/** Filename of the backing file if any. */
179	char *pszBackingFilename;
180	/** Offset of the filename in the image. */
181	uint32_t offBackingFilename;
182	/** Size of the backing filename excluding \0. */
183	uint32_t cbBackingFilename;
184
185	/** Size of the image, multiple of clusters. */
186	uint64_t cbImage;
187	/** Cluster size in bytes. */
188	uint32_t cbCluster;
189	/** Number of entries in the L1 and L2 table. */
190	uint32_t cTableEntries;
191	/** Size of an L1 or L2 table rounded to the next cluster size. */
192	uint32_t cbTable;
193	/** Pointer to the L1 table. */
194	uint64_t *paL1Table;
195	/** Offset of the L1 table. */
196	uint64_t offL1Table;
197
198	/** Offset mask for a cluster. */
199	uint64_t fOffsetMask;
200	/** L1 table mask to get the L1 index. */
201	uint64_t fL1Mask;
202	/** Number of bits to shift to get the L1 index. */
203	uint32_t cL1Shift;
204	/** L2 table mask to get the L2 index. */
205	uint64_t fL2Mask;
206	/** Number of bits to shift to get the L2 index. */
207	uint32_t cL2Shift;
208
209	/** Memory occupied by the L2 table cache. */
210	size_t cbL2Cache;
211	/** The sorted L2 entry list used for searching. */
212	RTLISTNODE ListSearch;
213	/** The LRU L2 entry list used for eviction. */
214	RTLISTNODE ListLru;
215
216	} QEDIMAGE, *PQEDIMAGE;
217
218	/**
219	* State of the async cluster allocation.
220	*/
221	typedef enum QEDCLUSTERASYNCALLOCSTATE
222	{
223	/** Invalid. */
224	QEDCLUSTERASYNCALLOCSTATE_INVALID = 0,
225	/** L2 table allocation. */
226	QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC,
227	/** Link L2 table into L1. */
228	QEDCLUSTERASYNCALLOCSTATE_L2_LINK,
229	/** Allocate user data cluster. */
230	QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC,
231	/** Link user data cluster. */
232	QEDCLUSTERASYNCALLOCSTATE_USER_LINK,
233	/** 32bit blowup. */
234	QEDCLUSTERASYNCALLOCSTATE_32BIT_HACK = 0x7fffffff
235	} QEDCLUSTERASYNCALLOCSTATE, *PQEDCLUSTERASYNCALLOCSTATE;
236
237	/**
238	* Data needed to track async cluster allocation.
239	*/
240	typedef struct QEDCLUSTERASYNCALLOC
241	{
242	/** The state of the cluster allocation. */
243	QEDCLUSTERASYNCALLOCSTATE enmAllocState;
244	/** Old image size to rollback in case of an error. */
245	uint64_t cbImageOld;
246	/** L1 index to link if any. */
247	uint32_t idxL1;
248	/** L2 index to link, required in any case. */
249	uint32_t idxL2;
250	/** Start offset of the allocated cluster. */
251	uint64_t offClusterNew;
252	/** L2 cache entry if a L2 table is allocated. */
253	PQEDL2CACHEENTRY pL2Entry;
254	/** Number of bytes to write. */
255	size_t cbToWrite;
256	} QEDCLUSTERASYNCALLOC, *PQEDCLUSTERASYNCALLOC;
257
258
259	/*********************************************************************************************************************************
260	* Static Variables *
261	*********************************************************************************************************************************/
262
263	/** NULL-terminated array of supported file extensions. */
264	static const VDFILEEXTENSION s_aQedFileExtensions[] =
265	{
266	{"qed", VDTYPE_HDD},
267	{NULL, VDTYPE_INVALID}
268	};
269
270
271	/*********************************************************************************************************************************
272	* Internal Functions *
273	*********************************************************************************************************************************/
274
275	/**
276	* Converts the image header to the host endianess and performs basic checks.
277	*
278	* @returns Whether the given header is valid or not.
279	* @param pHeader Pointer to the header to convert.
280	*/
281	static bool qedHdrConvertToHostEndianess(PQedHeader pHeader)
282	{
283	pHeader->u32Magic = RT_LE2H_U32(pHeader->u32Magic);
284	pHeader->u32ClusterSize = RT_LE2H_U32(pHeader->u32ClusterSize);
285	pHeader->u32TableSize = RT_LE2H_U32(pHeader->u32TableSize);
286	pHeader->u32HeaderSize = RT_LE2H_U32(pHeader->u32HeaderSize);
287	pHeader->u64FeatureFlags = RT_LE2H_U64(pHeader->u64FeatureFlags);
288	pHeader->u64CompatFeatureFlags = RT_LE2H_U64(pHeader->u64CompatFeatureFlags);
289	pHeader->u64AutoresetFeatureFlags = RT_LE2H_U64(pHeader->u64AutoresetFeatureFlags);
290	pHeader->u64OffL1Table = RT_LE2H_U64(pHeader->u64OffL1Table);
291	pHeader->u64Size = RT_LE2H_U64(pHeader->u64Size);
292	pHeader->u32OffBackingFilename = RT_LE2H_U32(pHeader->u32OffBackingFilename);
293	pHeader->u32BackingFilenameSize = RT_LE2H_U32(pHeader->u32BackingFilenameSize);
294
295	if (RT_UNLIKELY(pHeader->u32Magic != QED_MAGIC))
296	return false;
297	if (RT_UNLIKELY( pHeader->u32ClusterSize < QED_CLUSTER_SIZE_MIN
298	\|\| pHeader->u32ClusterSize > QED_CLUSTER_SIZE_MAX))
299	return false;
300	if (RT_UNLIKELY( pHeader->u32TableSize < QED_TABLE_SIZE_MIN
301	\|\| pHeader->u32TableSize > QED_TABLE_SIZE_MAX))
302	return false;
303	if (RT_UNLIKELY(pHeader->u64Size % 512 != 0))
304	return false;
305	if (RT_UNLIKELY( pHeader->u64FeatureFlags & QED_FEATURE_BACKING_FILE
306	&& ( pHeader->u32BackingFilenameSize == 0
307	\|\| pHeader->u32BackingFilenameSize == UINT32_MAX)))
308	return false;
309
310	return true;
311	}
312
313	/**
314	* Creates a QED header from the given image state.
315	*
316	* @returns nothing.
317	* @param pImage Image instance data.
318	* @param pHeader Pointer to the header to convert.
319	*/
320	static void qedHdrConvertFromHostEndianess(PQEDIMAGE pImage, PQedHeader pHeader)
321	{
322	pHeader->u32Magic = RT_H2LE_U32(QED_MAGIC);
323	pHeader->u32ClusterSize = RT_H2LE_U32(pImage->cbCluster);
324	pHeader->u32TableSize = RT_H2LE_U32(pImage->cbTable / pImage->cbCluster);
325	pHeader->u32HeaderSize = RT_H2LE_U32(1);
326	pHeader->u64FeatureFlags = RT_H2LE_U64(pImage->pszBackingFilename ? QED_FEATURE_BACKING_FILE : UINT64_C(0));
327	pHeader->u64CompatFeatureFlags = RT_H2LE_U64(UINT64_C(0));
328	pHeader->u64AutoresetFeatureFlags = RT_H2LE_U64(UINT64_C(0));
329	pHeader->u64OffL1Table = RT_H2LE_U64(pImage->offL1Table);
330	pHeader->u64Size = RT_H2LE_U64(pImage->cbSize);
331	pHeader->u32OffBackingFilename = RT_H2LE_U32(pImage->offBackingFilename);
332	pHeader->u32BackingFilenameSize = RT_H2LE_U32(pImage->cbBackingFilename);
333	}
334
335	/**
336	* Convert table entries from little endian to host endianess.
337	*
338	* @returns nothing.
339	* @param paTbl Pointer to the table.
340	* @param cEntries Number of entries in the table.
341	*/
342	static void qedTableConvertToHostEndianess(uint64_t *paTbl, uint32_t cEntries)
343	{
344	while(cEntries-- > 0)
345	{
346	paTbl = RT_LE2H_U64(paTbl);
347	paTbl++;
348	}
349	}
350
351	#if defined(RT_BIG_ENDIAN)
352	/**
353	* Convert table entries from host to little endian format.
354	*
355	* @returns nothing.
356	* @param paTblImg Pointer to the table which will store the little endian table.
357	* @param paTbl The source table to convert.
358	* @param cEntries Number of entries in the table.
359	*/
360	static void qedTableConvertFromHostEndianess(uint64_t paTblImg, uint64_t paTbl,
361	uint32_t cEntries)
362	{
363	while(cEntries-- > 0)
364	{
365	paTblImg = RT_H2LE_U64(paTbl);
366	paTbl++;
367	paTblImg++;
368	}
369	}
370	#endif
371
372	/**
373	* Creates the L2 table cache.
374	*
375	* @returns VBox status code.
376	* @param pImage The image instance data.
377	*/
378	static int qedL2TblCacheCreate(PQEDIMAGE pImage)
379	{
380	pImage->cbL2Cache = 0;
381	RTListInit(&pImage->ListSearch);
382	RTListInit(&pImage->ListLru);
383
384	return VINF_SUCCESS;
385	}
386
387	/**
388	* Destroys the L2 table cache.
389	*
390	* @returns nothing.
391	* @param pImage The image instance data.
392	*/
393	static void qedL2TblCacheDestroy(PQEDIMAGE pImage)
394	{
395	PQEDL2CACHEENTRY pL2Entry = NULL;
396	PQEDL2CACHEENTRY pL2Next = NULL;
397
398	RTListForEachSafe(&pImage->ListSearch, pL2Entry, pL2Next, QEDL2CACHEENTRY, NodeSearch)
399	{
400	Assert(!pL2Entry->cRefs);
401
402	RTListNodeRemove(&pL2Entry->NodeSearch);
403	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
404	RTMemFree(pL2Entry);
405	}
406
407	pImage->cbL2Cache = 0;
408	RTListInit(&pImage->ListSearch);
409	RTListInit(&pImage->ListLru);
410	}
411
412	/**
413	* Returns the L2 table matching the given offset or NULL if none could be found.
414	*
415	* @returns Pointer to the L2 table cache entry or NULL.
416	* @param pImage The image instance data.
417	* @param offL2Tbl Offset of the L2 table to search for.
418	*/
419	static PQEDL2CACHEENTRY qedL2TblCacheRetain(PQEDIMAGE pImage, uint64_t offL2Tbl)
420	{
421	PQEDL2CACHEENTRY pL2Entry = NULL;
422
423	RTListForEach(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch)
424	{
425	if (pL2Entry->offL2Tbl == offL2Tbl)
426	break;
427	}
428
429	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
430	{
431	/* Update LRU list. */
432	RTListNodeRemove(&pL2Entry->NodeLru);
433	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
434	pL2Entry->cRefs++;
435	return pL2Entry;
436	}
437	else
438	return NULL;
439	}
440
441	/**
442	* Releases a L2 table cache entry.
443	*
444	* @returns nothing.
445	* @param pL2Entry The L2 cache entry.
446	*/
447	static void qedL2TblCacheEntryRelease(PQEDL2CACHEENTRY pL2Entry)
448	{
449	Assert(pL2Entry->cRefs > 0);
450	pL2Entry->cRefs--;
451	}
452
453	/**
454	* Allocates a new L2 table from the cache evicting old entries if required.
455	*
456	* @returns Pointer to the L2 cache entry or NULL.
457	* @param pImage The image instance data.
458	*/
459	static PQEDL2CACHEENTRY qedL2TblCacheEntryAlloc(PQEDIMAGE pImage)
460	{
461	PQEDL2CACHEENTRY pL2Entry = NULL;
462
463	if (pImage->cbL2Cache + pImage->cbTable <= QED_L2_CACHE_MEMORY_MAX)
464	{
465	/* Add a new entry. */
466	pL2Entry = (PQEDL2CACHEENTRY)RTMemAllocZ(sizeof(QEDL2CACHEENTRY));
467	if (pL2Entry)
468	{
469	pL2Entry->paL2Tbl = (uint64_t *)RTMemPageAllocZ(pImage->cbTable);
470	if (RT_UNLIKELY(!pL2Entry->paL2Tbl))
471	{
472	RTMemFree(pL2Entry);
473	pL2Entry = NULL;
474	}
475	else
476	{
477	pL2Entry->cRefs = 1;
478	pImage->cbL2Cache += pImage->cbTable;
479	}
480	}
481	}
482	else
483	{
484	/* Evict the last not in use entry and use it */
485	Assert(!RTListIsEmpty(&pImage->ListLru));
486
487	RTListForEachReverse(&pImage->ListLru, pL2Entry, QEDL2CACHEENTRY, NodeLru)
488	{
489	if (!pL2Entry->cRefs)
490	break;
491	}
492
493	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
494	{
495	RTListNodeRemove(&pL2Entry->NodeSearch);
496	RTListNodeRemove(&pL2Entry->NodeLru);
497	pL2Entry->offL2Tbl = 0;
498	pL2Entry->cRefs = 1;
499	}
500	else
501	pL2Entry = NULL;
502	}
503
504	return pL2Entry;
505	}
506
507	/**
508	* Frees a L2 table cache entry.
509	*
510	* @returns nothing.
511	* @param pImage The image instance data.
512	* @param pL2Entry The L2 cache entry to free.
513	*/
514	static void qedL2TblCacheEntryFree(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
515	{
516	Assert(!pL2Entry->cRefs);
517	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
518	RTMemFree(pL2Entry);
519
520	pImage->cbL2Cache -= pImage->cbTable;
521	}
522
523	/**
524	* Inserts an entry in the L2 table cache.
525	*
526	* @returns nothing.
527	* @param pImage The image instance data.
528	* @param pL2Entry The L2 cache entry to insert.
529	*/
530	static void qedL2TblCacheEntryInsert(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
531	{
532	PQEDL2CACHEENTRY pIt = NULL;
533
534	Assert(pL2Entry->offL2Tbl > 0);
535
536	/* Insert at the top of the LRU list. */
537	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
538
539	if (RTListIsEmpty(&pImage->ListSearch))
540	{
541	RTListAppend(&pImage->ListSearch, &pL2Entry->NodeSearch);
542	}
543	else
544	{
545	/* Insert into search list. */
546	pIt = RTListGetFirst(&pImage->ListSearch, QEDL2CACHEENTRY, NodeSearch);
547	if (pIt->offL2Tbl > pL2Entry->offL2Tbl)
548	RTListPrepend(&pImage->ListSearch, &pL2Entry->NodeSearch);
549	else
550	{
551	bool fInserted = false;
552
553	RTListForEach(&pImage->ListSearch, pIt, QEDL2CACHEENTRY, NodeSearch)
554	{
555	Assert(pIt->offL2Tbl != pL2Entry->offL2Tbl);
556	if (pIt->offL2Tbl < pL2Entry->offL2Tbl)
557	{
558	RTListNodeInsertAfter(&pIt->NodeSearch, &pL2Entry->NodeSearch);
559	fInserted = true;
560	break;
561	}
562	}
563	Assert(fInserted);
564	}
565	}
566	}
567
568	/**
569	* Fetches the L2 from the given offset trying the LRU cache first and
570	* reading it from the image after a cache miss - version for async I/O.
571	*
572	* @returns VBox status code.
573	* @param pImage Image instance data.
574	* @param pIoCtx The I/O context.
575	* @param offL2Tbl The offset of the L2 table in the image.
576	* @param ppL2Entry Where to store the L2 table on success.
577	*/
578	static int qedL2TblCacheFetchAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
579	uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
580	{
581	int rc = VINF_SUCCESS;
582
583	/* Try to fetch the L2 table from the cache first. */
584	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
585	if (!pL2Entry)
586	{
587	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
588
589	if (pL2Entry)
590	{
591	/* Read from the image. */
592	PVDMETAXFER pMetaXfer;
593
594	pL2Entry->offL2Tbl = offL2Tbl;
595	rc = vdIfIoIntFileReadMeta(pImage->pIfIo, pImage->pStorage,
596	offL2Tbl, pL2Entry->paL2Tbl,
597	pImage->cbTable, pIoCtx,
598	&pMetaXfer, NULL, NULL);
599	if (RT_SUCCESS(rc))
600	{
601	vdIfIoIntMetaXferRelease(pImage->pIfIo, pMetaXfer);
602	#if defined(RT_BIG_ENDIAN)
603	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
604	#endif
605	qedL2TblCacheEntryInsert(pImage, pL2Entry);
606	}
607	else
608	{
609	qedL2TblCacheEntryRelease(pL2Entry);
610	qedL2TblCacheEntryFree(pImage, pL2Entry);
611	}
612	}
613	else
614	rc = VERR_NO_MEMORY;
615	}
616
617	if (RT_SUCCESS(rc))
618	*ppL2Entry = pL2Entry;
619
620	return rc;
621	}
622
623	/**
624	* Return power of 2 or 0 if num error.
625	*
626	* @returns The power of 2 or 0 if the given number is not a power of 2.
627	* @param u32 The number.
628	*/
629	static uint32_t qedGetPowerOfTwo(uint32_t u32)
630	{
631	if (u32 == 0)
632	return 0;
633	uint32_t uPower2 = 0;
634	while ((u32 & 1) == 0)
635	{
636	u32 >>= 1;
637	uPower2++;
638	}
639	return u32 == 1 ? uPower2 : 0;
640	}
641
642	/**
643	* Sets the L1, L2 and offset bitmasks and L1 and L2 bit shift members.
644	*
645	* @returns nothing.
646	* @param pImage The image instance data.
647	*/
648	static void qedTableMasksInit(PQEDIMAGE pImage)
649	{
650	uint32_t cClusterBits, cTableBits;
651
652	cClusterBits = qedGetPowerOfTwo(pImage->cbCluster);
653	cTableBits = qedGetPowerOfTwo(pImage->cTableEntries);
654
655	Assert(cClusterBits + 2 * cTableBits <= 64);
656
657	pImage->fOffsetMask = ((uint64_t)pImage->cbCluster - 1);
658	pImage->fL2Mask = ((uint64_t)pImage->cTableEntries - 1) << cClusterBits;
659	pImage->cL2Shift = cClusterBits;
660	pImage->fL1Mask = ((uint64_t)pImage->cTableEntries - 1) << (cClusterBits + cTableBits);
661	pImage->cL1Shift = cClusterBits + cTableBits;
662	}
663
664	/**
665	* Converts a given logical offset into the
666	*
667	* @returns nothing.
668	* @param pImage The image instance data.
669	* @param off The logical offset to convert.
670	* @param pidxL1 Where to store the index in the L1 table on success.
671	* @param pidxL2 Where to store the index in the L2 table on success.
672	* @param poffCluster Where to store the offset in the cluster on success.
673	*/
674	DECLINLINE(void) qedConvertLogicalOffset(PQEDIMAGE pImage, uint64_t off, uint32_t *pidxL1,
675	uint32_t pidxL2, uint32_t poffCluster)
676	{
677	AssertPtr(pidxL1);
678	AssertPtr(pidxL2);
679	AssertPtr(poffCluster);
680
681	*poffCluster = off & pImage->fOffsetMask;
682	*pidxL1 = (off & pImage->fL1Mask) >> pImage->cL1Shift;
683	*pidxL2 = (off & pImage->fL2Mask) >> pImage->cL2Shift;
684	}
685
686	/**
687	* Converts Cluster size to a byte size.
688	*
689	* @returns Number of bytes derived from the given number of clusters.
690	* @param pImage The image instance data.
691	* @param cClusters The clusters to convert.
692	*/
693	DECLINLINE(uint64_t) qedCluster2Byte(PQEDIMAGE pImage, uint64_t cClusters)
694	{
695	return cClusters * pImage->cbCluster;
696	}
697
698	/**
699	* Converts number of bytes to cluster size rounding to the next cluster.
700	*
701	* @returns Number of bytes derived from the given number of clusters.
702	* @param pImage The image instance data.
703	* @param cb Number of bytes to convert.
704	*/
705	DECLINLINE(uint64_t) qedByte2Cluster(PQEDIMAGE pImage, uint64_t cb)
706	{
707	return cb / pImage->cbCluster + (cb % pImage->cbCluster ? 1 : 0);
708	}
709
710	/**
711	* Allocates a new cluster in the image.
712	*
713	* @returns The start offset of the new cluster in the image.
714	* @param pImage The image instance data.
715	* @param cCLusters Number of clusters to allocate.
716	*/
717	DECLINLINE(uint64_t) qedClusterAllocate(PQEDIMAGE pImage, uint32_t cClusters)
718	{
719	uint64_t offCluster;
720
721	offCluster = pImage->cbImage;
722	pImage->cbImage += cClusters*pImage->cbCluster;
723
724	return offCluster;
725	}
726
727	/**
728	* Returns the real image offset for a given cluster or an error if the cluster is not
729	* yet allocated.
730	*
731	* @returns VBox status code.
732	* VERR_VD_BLOCK_FREE if the cluster is not yet allocated.
733	* @param pImage The image instance data.
734	* @param pIoCtx The I/O context.
735	* @param idxL1 The L1 index.
736	* @param idxL2 The L2 index.
737	* @param offCluster Offset inside the cluster.
738	* @param poffImage Where to store the image offset on success;
739	*/
740	static int qedConvertToImageOffset(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
741	uint32_t idxL1, uint32_t idxL2,
742	uint32_t offCluster, uint64_t *poffImage)
743	{
744	int rc = VERR_VD_BLOCK_FREE;
745
746	AssertReturn(idxL1 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
747	AssertReturn(idxL2 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
748
749	if (pImage->paL1Table[idxL1])
750	{
751	PQEDL2CACHEENTRY pL2Entry;
752
753	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
754	&pL2Entry);
755	if (RT_SUCCESS(rc))
756	{
757	/* Get real file offset. */
758	if (pL2Entry->paL2Tbl[idxL2])
759	*poffImage = pL2Entry->paL2Tbl[idxL2] + offCluster;
760	else
761	rc = VERR_VD_BLOCK_FREE;
762
763	qedL2TblCacheEntryRelease(pL2Entry);
764	}
765	}
766
767	return rc;
768	}
769
770	/**
771	* Write the given table to image converting to the image endianess if required.
772	*
773	* @returns VBox status code.
774	* @param pImage The image instance data.
775	* @param pIoCtx The I/O context.
776	* @param offTbl The offset the table should be written to.
777	* @param paTbl The table to write.
778	* @param pfnComplete Callback called when the write completes.
779	* @param pvUser Opaque user data to pass in the completion callback.
780	*/
781	static int qedTblWrite(PQEDIMAGE pImage, PVDIOCTX pIoCtx, uint64_t offTbl, uint64_t *paTbl,
782	PFNVDXFERCOMPLETED pfnComplete, void *pvUser)
783	{
784	int rc = VINF_SUCCESS;
785
786	#if defined(RT_BIG_ENDIAN)
787	uint64_t paTblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
788	if (paTblImg)
789	{
790	qedTableConvertFromHostEndianess(paTblImg, paTbl,
791	pImage->cTableEntries);
792	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
793	offTbl, paTblImg, pImage->cbTable,
794	pIoCtx, pfnComplete, pvUser);
795	RTMemFree(paTblImg);
796	}
797	else
798	rc = VERR_NO_MEMORY;
799	#else
800	/* Write table directly. */
801	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
802	offTbl, paTbl, pImage->cbTable, pIoCtx,
803	pfnComplete, pvUser);
804	#endif
805
806	return rc;
807	}
808
809	/**
810	* Internal. Flush image data to disk.
811	*/
812	static int qedFlushImage(PQEDIMAGE pImage)
813	{
814	int rc = VINF_SUCCESS;
815
816	if ( pImage->pStorage
817	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
818	{
819	QedHeader Header;
820
821	Assert(!(pImage->cbTable % pImage->cbCluster));
822	#if defined(RT_BIG_ENDIAN)
823	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
824	if (paL1TblImg)
825	{
826	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
827	pImage->cTableEntries);
828	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
829	pImage->offL1Table, paL1TblImg,
830	pImage->cbTable);
831	RTMemFree(paL1TblImg);
832	}
833	else
834	rc = VERR_NO_MEMORY;
835	#else
836	/* Write L1 table directly. */
837	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, pImage->offL1Table,
838	pImage->paL1Table, pImage->cbTable);
839	#endif
840	if (RT_SUCCESS(rc))
841	{
842	/* Write header. */
843	qedHdrConvertFromHostEndianess(pImage, &Header);
844	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, 0, &Header,
845	sizeof(Header));
846	if (RT_SUCCESS(rc))
847	rc = vdIfIoIntFileFlushSync(pImage->pIfIo, pImage->pStorage);
848	}
849	}
850
851	return rc;
852	}
853
854	/**
855	* Flush image data to disk - version for async I/O.
856	*
857	* @returns VBox status code.
858	* @param pImage The image instance data.
859	* @param pIoCtx The I/o context
860	*/
861	static int qedFlushImageAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx)
862	{
863	int rc = VINF_SUCCESS;
864
865	if ( pImage->pStorage
866	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
867	{
868	QedHeader Header;
869
870	Assert(!(pImage->cbTable % pImage->cbCluster));
871	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
872	NULL, NULL);
873	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
874	{
875	/* Write header. */
876	qedHdrConvertFromHostEndianess(pImage, &Header);
877	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
878	0, &Header, sizeof(Header),
879	pIoCtx, NULL, NULL);
880	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
881	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
882	pIoCtx, NULL, NULL);
883	}
884	}
885
886	return rc;
887	}
888
889	/**
890	* Checks whether the given cluster offset is valid.
891	*
892	* @returns Whether the given cluster offset is valid.
893	* @param offCluster The table offset to check.
894	* @param cbFile The real file size of the image.
895	* @param cbCluster The cluster size in bytes.
896	*/
897	DECLINLINE(bool) qedIsClusterOffsetValid(uint64_t offCluster, uint64_t cbFile, size_t cbCluster)
898	{
899	return (offCluster <= cbFile - cbCluster)
900	&& !(offCluster & (cbCluster - 1));
901	}
902
903	/**
904	* Checks whether the given table offset is valid.
905	*
906	* @returns Whether the given table offset is valid.
907	* @param offTbl The table offset to check.
908	* @param cbFile The real file size of the image.
909	* @param cbTable The table size in bytes.
910	* @param cbCluster The cluster size in bytes.
911	*/
912	DECLINLINE(bool) qedIsTblOffsetValid(uint64_t offTbl, uint64_t cbFile, size_t cbTable, size_t cbCluster)
913	{
914	return (offTbl <= cbFile - cbTable)
915	&& !(offTbl & (cbCluster - 1));
916	}
917
918	/**
919	* Sets the specified range in the cluster bitmap checking whether any of the clusters is already
920	* used before.
921	*
922	* @returns Whether the range was clear and is set now.
923	* @param pvClusterBitmap The cluster bitmap to use.
924	* @param offClusterStart The first cluster to check and set.
925	* @param offClusterEnd The first cluster to not check and set anymore.
926	*/
927	static bool qedClusterBitmapCheckAndSet(void *pvClusterBitmap, uint32_t offClusterStart, uint32_t offClusterEnd)
928	{
929	for (uint32_t offCluster = offClusterStart; offCluster < offClusterEnd; offCluster++)
930	if (ASMBitTest(pvClusterBitmap, offCluster))
931	return false;
932
933	ASMBitSetRange(pvClusterBitmap, offClusterStart, offClusterEnd);
934	return true;
935	}
936
937	/**
938	* Checks the given image for consistency, usually called when the
939	* QED_FEATURE_NEED_CHECK bit is set.
940	*
941	* @returns VBox status code.
942	* @retval VINF_SUCCESS when the image can be accessed.
943	* @param pImage The image instance data.
944	* @param pHeader The header to use for checking.
945	*
946	* @note It is not required that the image state is fully initialized Only
947	* The I/O interface and storage handle need to be valid.
948	* @note The header must be converted to the host CPU endian format already
949	* and should be validated already.
950	*/
951	static int qedCheckImage(PQEDIMAGE pImage, PQedHeader pHeader)
952	{
953	uint64_t cbFile;
954	uint32_t cbTable;
955	uint32_t cTableEntries;
956	uint64_t *paL1Tbl = NULL;
957	uint64_t *paL2Tbl = NULL;
958	void *pvClusterBitmap = NULL;
959	uint32_t offClusterStart;
960	int rc = VINF_SUCCESS;
961
962	pImage->cbCluster = pHeader->u32ClusterSize;
963	cbTable = pHeader->u32TableSize * pHeader->u32ClusterSize;
964	cTableEntries = cbTable / sizeof(uint64_t);
965
966	do
967	{
968	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
969	if (RT_FAILURE(rc))
970	{
971	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
972	N_("Qed: Querying the file size of image '%s' failed"),
973	pImage->pszFilename);
974	break;
975	}
976
977	/* Allocate L1 table. */
978	paL1Tbl = (uint64_t *)RTMemAllocZ(cbTable);
979	if (!paL1Tbl)
980	{
981	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
982	N_("Qed: Allocating memory for the L1 table for image '%s' failed"),
983	pImage->pszFilename);
984	break;
985	}
986
987	paL2Tbl = (uint64_t *)RTMemAllocZ(cbTable);
988	if (!paL2Tbl)
989	{
990	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
991	N_("Qed: Allocating memory for the L2 table for image '%s' failed"),
992	pImage->pszFilename);
993	break;
994	}
995
996	pvClusterBitmap = RTMemAllocZ(cbFile / pHeader->u32ClusterSize / 8);
997	if (!pvClusterBitmap)
998	{
999	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1000	N_("Qed: Allocating memory for the cluster bitmap for image '%s' failed"),
1001	pImage->pszFilename);
1002	break;
1003	}
1004
1005	/* Validate L1 table offset. */
1006	if (!qedIsTblOffsetValid(pHeader->u64OffL1Table, cbFile, cbTable, pHeader->u32ClusterSize))
1007	{
1008	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1009	N_("Qed: L1 table offset of image '%s' is corrupt (%llu)"),
1010	pImage->pszFilename, pHeader->u64OffL1Table);
1011	break;
1012	}
1013
1014	/* Read L1 table. */
1015	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1016	pHeader->u64OffL1Table, paL1Tbl, cbTable);
1017	if (RT_FAILURE(rc))
1018	{
1019	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1020	N_("Qed: Reading the L1 table from image '%s' failed"),
1021	pImage->pszFilename);
1022	break;
1023	}
1024
1025	/* Mark the L1 table in cluster bitmap. */
1026	ASMBitSet(pvClusterBitmap, 0); /* Header is always in cluster 0. */
1027	offClusterStart = qedByte2Cluster(pImage, pHeader->u64OffL1Table);
1028	bool fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1029	Assert(fSet);
1030
1031	/* Scan the L1 and L2 tables for invalid entries. */
1032	qedTableConvertToHostEndianess(paL1Tbl, cTableEntries);
1033
1034	for (unsigned iL1 = 0; iL1 < cTableEntries; iL1++)
1035	{
1036	if (!paL1Tbl[iL1])
1037	continue; /* Skip unallocated clusters. */
1038
1039	if (!qedIsTblOffsetValid(paL1Tbl[iL1], cbFile, cbTable, pHeader->u32ClusterSize))
1040	{
1041	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1042	N_("Qed: Entry %d of the L1 table from image '%s' is invalid (%llu)"),
1043	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1044	break;
1045	}
1046
1047	/* Now check that the clusters are not allocated already. */
1048	offClusterStart = qedByte2Cluster(pImage, paL1Tbl[iL1]);
1049	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1050	if (!fSet)
1051	{
1052	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1053	N_("Qed: Entry %d of the L1 table from image '%s' points to a already used cluster (%llu)"),
1054	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1055	break;
1056	}
1057
1058	/* Read the linked L2 table and check it. */
1059	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1060	paL1Tbl[iL1], paL2Tbl, cbTable);
1061	if (RT_FAILURE(rc))
1062	{
1063	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1064	N_("Qed: Reading the L2 table from image '%s' failed"),
1065	pImage->pszFilename);
1066	break;
1067	}
1068
1069	/* Check all L2 entries. */
1070	for (unsigned iL2 = 0; iL2 < cTableEntries; iL2++)
1071	{
1072	if (paL2Tbl[iL2])
1073	continue; /* Skip unallocated clusters. */
1074
1075	if (!qedIsClusterOffsetValid(paL2Tbl[iL2], cbFile, pHeader->u32ClusterSize))
1076	{
1077	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1078	N_("Qed: Entry %d of the L2 table from image '%s' is invalid (%llu)"),
1079	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1080	break;
1081	}
1082
1083	/* Now check that the clusters are not allocated already. */
1084	offClusterStart = qedByte2Cluster(pImage, paL2Tbl[iL2]);
1085	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + 1);
1086	if (!fSet)
1087	{
1088	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1089	N_("Qed: Entry %d of the L2 table from image '%s' points to a already used cluster (%llu)"),
1090	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1091	break;
1092	}
1093	}
1094	}
1095	} while(0);
1096
1097	if (paL1Tbl)
1098	RTMemFree(paL1Tbl);
1099	if (paL2Tbl)
1100	RTMemFree(paL2Tbl);
1101	if (pvClusterBitmap)
1102	RTMemFree(pvClusterBitmap);
1103
1104	return rc;
1105	}
1106
1107	/**
1108	* Internal. Free all allocated space for representing an image except pImage,
1109	* and optionally delete the image from disk.
1110	*/
1111	static int qedFreeImage(PQEDIMAGE pImage, bool fDelete)
1112	{
1113	int rc = VINF_SUCCESS;
1114
1115	/* Freeing a never allocated image (e.g. because the open failed) is
1116	* not signalled as an error. After all nothing bad happens. */
1117	if (pImage)
1118	{
1119	if (pImage->pStorage)
1120	{
1121	/* No point updating the file that is deleted anyway. */
1122	if (!fDelete)
1123	qedFlushImage(pImage);
1124
1125	rc = vdIfIoIntFileClose(pImage->pIfIo, pImage->pStorage);
1126	pImage->pStorage = NULL;
1127	}
1128
1129	if (pImage->paL1Table)
1130	RTMemFree(pImage->paL1Table);
1131
1132	if (pImage->pszBackingFilename)
1133	{
1134	RTMemFree(pImage->pszBackingFilename);
1135	pImage->pszBackingFilename = NULL;
1136	}
1137
1138	qedL2TblCacheDestroy(pImage);
1139
1140	if (fDelete && pImage->pszFilename)
1141	vdIfIoIntFileDelete(pImage->pIfIo, pImage->pszFilename);
1142	}
1143
1144	LogFlowFunc(("returns %Rrc\n", rc));
1145	return rc;
1146	}
1147
1148	/**
1149	* Internal: Open an image, constructing all necessary data structures.
1150	*/
1151	static int qedOpenImage(PQEDIMAGE pImage, unsigned uOpenFlags)
1152	{
1153	int rc;
1154
1155	pImage->uOpenFlags = uOpenFlags;
1156
1157	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1158	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1159	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1160
1161	/*
1162	* Create the L2 cache before opening the image so we can call qedFreeImage()
1163	* even if opening the image file fails.
1164	*/
1165	rc = qedL2TblCacheCreate(pImage);
1166	if (RT_FAILURE(rc))
1167	{
1168	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1169	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1170	pImage->pszFilename);
1171
1172	goto out;
1173	}
1174
1175	/*
1176	* Open the image.
1177	*/
1178	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename,
1179	VDOpenFlagsToFileOpenFlags(uOpenFlags,
1180	false /* fCreate */),
1181	&pImage->pStorage);
1182	if (RT_FAILURE(rc))
1183	{
1184	/* Do NOT signal an appropriate error here, as the VD layer has the
1185	* choice of retrying the open if it failed. */
1186	goto out;
1187	}
1188
1189	uint64_t cbFile;
1190	QedHeader Header;
1191	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1192	if (RT_FAILURE(rc))
1193	goto out;
1194	if (cbFile > sizeof(Header))
1195	{
1196	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, 0, &Header, sizeof(Header));
1197	if ( RT_SUCCESS(rc)
1198	&& qedHdrConvertToHostEndianess(&Header))
1199	{
1200	if ( !(Header.u64FeatureFlags & ~QED_FEATURE_MASK)
1201	&& !(Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE_NO_PROBE))
1202	{
1203	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1204	{
1205	/* Image needs checking. */
1206	if (!(uOpenFlags & VD_OPEN_FLAGS_READONLY))
1207	rc = qedCheckImage(pImage, &Header);
1208	else
1209	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1210	N_("Qed: Image '%s' needs checking but is opened readonly"),
1211	pImage->pszFilename);
1212	}
1213
1214	if ( RT_SUCCESS(rc)
1215	&& (Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE))
1216	{
1217	/* Load backing filename from image. */
1218	pImage->pszBackingFilename = (char )RTMemAllocZ(Header.u32BackingFilenameSize + 1); / +1 for \0 terminator. */
1219	if (pImage->pszBackingFilename)
1220	{
1221	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1222	pImage->offBackingFilename = Header.u32OffBackingFilename;
1223	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1224	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1225	Header.u32BackingFilenameSize);
1226	}
1227	else
1228	rc = VERR_NO_MEMORY;
1229	}
1230
1231	if (RT_SUCCESS(rc))
1232	{
1233	pImage->cbImage = cbFile;
1234	pImage->cbCluster = Header.u32ClusterSize;
1235	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1236	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1237	pImage->offL1Table = Header.u64OffL1Table;
1238	pImage->cbSize = Header.u64Size;
1239	qedTableMasksInit(pImage);
1240
1241	/* Allocate L1 table. */
1242	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1243	if (pImage->paL1Table)
1244	{
1245	/* Read from the image. */
1246	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1247	pImage->offL1Table, pImage->paL1Table,
1248	pImage->cbTable);
1249	if (RT_SUCCESS(rc))
1250	{
1251	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1252
1253	/* If the consistency check succeeded, clear the flag by flushing the image. */
1254	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1255	rc = qedFlushImage(pImage);
1256	}
1257	else
1258	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1259	N_("Qed: Reading the L1 table for image '%s' failed"),
1260	pImage->pszFilename);
1261	}
1262	else
1263	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1264	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1265	pImage->pszFilename);
1266	}
1267	}
1268	else
1269	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1270	N_("Qed: The image '%s' makes use of unsupported features"),
1271	pImage->pszFilename);
1272	}
1273	else if (RT_SUCCESS(rc))
1274	rc = VERR_VD_GEN_INVALID_HEADER;
1275	}
1276	else
1277	rc = VERR_VD_GEN_INVALID_HEADER;
1278
1279	out:
1280	if (RT_FAILURE(rc))
1281	qedFreeImage(pImage, false);
1282	return rc;
1283	}
1284
1285	/**
1286	* Internal: Create a qed image.
1287	*/
1288	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1289	unsigned uImageFlags, const char *pszComment,
1290	PCVDGEOMETRY pPCHSGeometry,
1291	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1292	PFNVDPROGRESS pfnProgress, void *pvUser,
1293	unsigned uPercentStart, unsigned uPercentSpan)
1294	{
1295	RT_NOREF1(pszComment);
1296	int rc;
1297	int32_t fOpen;
1298
1299	if (uImageFlags & VD_IMAGE_FLAGS_FIXED)
1300	{
1301	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1302	goto out;
1303	}
1304
1305	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1306	pImage->uImageFlags = uImageFlags;
1307	pImage->PCHSGeometry = *pPCHSGeometry;
1308	pImage->LCHSGeometry = *pLCHSGeometry;
1309
1310	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1311	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1312	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1313
1314	/* Create image file. */
1315	fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1316	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1317	if (RT_FAILURE(rc))
1318	{
1319	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1320	goto out;
1321	}
1322
1323	/* Init image state. */
1324	pImage->cbSize = cbSize;
1325	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1326	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1327	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1328	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1329	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1330	pImage->cbBackingFilename = 0;
1331	pImage->offBackingFilename = 0;
1332	qedTableMasksInit(pImage);
1333
1334	/* Init L1 table. */
1335	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1336	if (!pImage->paL1Table)
1337	{
1338	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1339	pImage->pszFilename);
1340	goto out;
1341	}
1342
1343	rc = qedL2TblCacheCreate(pImage);
1344	if (RT_FAILURE(rc))
1345	{
1346	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1347	pImage->pszFilename);
1348	goto out;
1349	}
1350
1351	if (RT_SUCCESS(rc) && pfnProgress)
1352	pfnProgress(pvUser, uPercentStart + uPercentSpan * 98 / 100);
1353
1354	rc = qedFlushImage(pImage);
1355
1356	out:
1357	if (RT_SUCCESS(rc) && pfnProgress)
1358	pfnProgress(pvUser, uPercentStart + uPercentSpan);
1359
1360	if (RT_FAILURE(rc))
1361	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1362	return rc;
1363	}
1364
1365	/**
1366	* Rollback anything done during async cluster allocation.
1367	*
1368	* @returns VBox status code.
1369	* @param pImage The image instance data.
1370	* @param pIoCtx The I/O context.
1371	* @param pClusterAlloc The cluster allocation to rollback.
1372	*/
1373	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1374	{
1375	RT_NOREF1(pIoCtx);
1376	int rc = VINF_SUCCESS;
1377
1378	switch (pClusterAlloc->enmAllocState)
1379	{
1380	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1381	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1382	{
1383	/* Revert the L1 table entry */
1384	pImage->paL1Table[pClusterAlloc->idxL1] = 0;
1385
1386	/* Assumption right now is that the L1 table is not modified on storage if the link fails. */
1387	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1388	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1389	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1390	break;
1391	}
1392	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1393	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1394	{
1395	/* Assumption right now is that the L2 table is not modified if the link fails. */
1396	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = 0;
1397	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1398	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1399	break;
1400	}
1401	default:
1402	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1403	rc = VERR_INVALID_STATE;
1404	}
1405
1406	RTMemFree(pClusterAlloc);
1407	return rc;
1408	}
1409
1410	/**
1411	* Updates the state of the async cluster allocation.
1412	*
1413	* @returns VBox status code.
1414	* @param pBackendData The opaque backend data.
1415	* @param pIoCtx I/O context associated with this request.
1416	* @param pvUser Opaque user data passed during a read/write request.
1417	* @param rcReq Status code for the completed request.
1418	*/
1419	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1420	{
1421	int rc = VINF_SUCCESS;
1422	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1423	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1424
1425	if (RT_FAILURE(rcReq))
1426	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1427
1428	AssertPtr(pClusterAlloc->pL2Entry);
1429
1430	switch (pClusterAlloc->enmAllocState)
1431	{
1432	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1433	{
1434	/* Update the link in the in memory L1 table now. */
1435	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1436
1437	/* Update the link in the on disk L1 table now. */
1438	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1439	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1440	qedAsyncClusterAllocUpdate, pClusterAlloc);
1441	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1442	break;
1443	else if (RT_FAILURE(rc))
1444	{
1445	/* Rollback. */
1446	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1447	break;
1448	}
1449	/* Success, fall through. */
1450	}
1451	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1452	{
1453	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1454	uint64_t offData = qedClusterAllocate(pImage, 1);
1455
1456	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1457
1458	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1459	pClusterAlloc->cbImageOld = offData;
1460	pClusterAlloc->offClusterNew = offData;
1461
1462	/* Write data. */
1463	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1464	offData, pIoCtx, pClusterAlloc->cbToWrite,
1465	qedAsyncClusterAllocUpdate, pClusterAlloc);
1466	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1467	break;
1468	else if (RT_FAILURE(rc))
1469	{
1470	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1471	RTMemFree(pClusterAlloc);
1472	break;
1473	}
1474	}
1475	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1476	{
1477	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1478	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1479
1480	/* Link L2 table and update it. */
1481	rc = qedTblWrite(pImage, pIoCtx, pImage->paL1Table[pClusterAlloc->idxL1],
1482	pClusterAlloc->pL2Entry->paL2Tbl,
1483	qedAsyncClusterAllocUpdate, pClusterAlloc);
1484	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1485	break;
1486	else if (RT_FAILURE(rc))
1487	{
1488	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1489	RTMemFree(pClusterAlloc);
1490	break;
1491	}
1492	}
1493	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1494	{
1495	/* Everything done without errors, signal completion. */
1496	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1497	RTMemFree(pClusterAlloc);
1498	rc = VINF_SUCCESS;
1499	break;
1500	}
1501	default:
1502	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1503	pClusterAlloc->enmAllocState));
1504	}
1505
1506	return rc;
1507	}
1508
1509	/** @copydoc VBOXHDDBACKEND::pfnCheckIfValid */
1510	static DECLCALLBACK(int) qedCheckIfValid(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1511	PVDINTERFACE pVDIfsImage, VDTYPE *penmType)
1512	{
1513	RT_NOREF1(pVDIfsDisk);
1514	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1515	PVDIOSTORAGE pStorage = NULL;
1516	uint64_t cbFile;
1517	int rc = VINF_SUCCESS;
1518
1519	/* Get I/O interface. */
1520	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1521	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1522
1523	if ( !VALID_PTR(pszFilename)
1524	\|\| !*pszFilename)
1525	{
1526	rc = VERR_INVALID_PARAMETER;
1527	goto out;
1528	}
1529
1530	/*
1531	* Open the file and read the footer.
1532	*/
1533	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1534	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1535	false /* fCreate */),
1536	&pStorage);
1537	if (RT_SUCCESS(rc))
1538	{
1539	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1540	if ( RT_SUCCESS(rc)
1541	&& cbFile > sizeof(QedHeader))
1542	{
1543	QedHeader Header;
1544
1545	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1546	if ( RT_SUCCESS(rc)
1547	&& qedHdrConvertToHostEndianess(&Header))
1548	{
1549	*penmType = VDTYPE_HDD;
1550	rc = VINF_SUCCESS;
1551	}
1552	else
1553	rc = VERR_VD_GEN_INVALID_HEADER;
1554	}
1555	else
1556	rc = VERR_VD_GEN_INVALID_HEADER;
1557	}
1558
1559	if (pStorage)
1560	vdIfIoIntFileClose(pIfIo, pStorage);
1561
1562	out:
1563	LogFlowFunc(("returns %Rrc\n", rc));
1564	return rc;
1565	}
1566
1567	/** @copydoc VBOXHDDBACKEND::pfnOpen */
1568	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1569	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1570	VDTYPE enmType, void **ppBackendData)
1571	{
1572	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n", pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1573	int rc;
1574	PQEDIMAGE pImage;
1575
1576	NOREF(enmType); /*< @todo r=klaus make use of the type info. /
1577
1578	/* Check open flags. All valid flags are supported. */
1579	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1580	{
1581	rc = VERR_INVALID_PARAMETER;
1582	goto out;
1583	}
1584
1585	/* Check remaining arguments. */
1586	if ( !VALID_PTR(pszFilename)
1587	\|\| !*pszFilename)
1588	{
1589	rc = VERR_INVALID_PARAMETER;
1590	goto out;
1591	}
1592
1593
1594	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1595	if (!pImage)
1596	{
1597	rc = VERR_NO_MEMORY;
1598	goto out;
1599	}
1600	pImage->pszFilename = pszFilename;
1601	pImage->pStorage = NULL;
1602	pImage->pVDIfsDisk = pVDIfsDisk;
1603	pImage->pVDIfsImage = pVDIfsImage;
1604
1605	rc = qedOpenImage(pImage, uOpenFlags);
1606	if (RT_SUCCESS(rc))
1607	*ppBackendData = pImage;
1608	else
1609	RTMemFree(pImage);
1610
1611	out:
1612	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1613	return rc;
1614	}
1615
1616	/** @copydoc VBOXHDDBACKEND::pfnCreate */
1617	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1618	unsigned uImageFlags, const char *pszComment,
1619	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1620	PCRTUUID pUuid, unsigned uOpenFlags,
1621	unsigned uPercentStart, unsigned uPercentSpan,
1622	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1623	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1624	void **ppBackendData)
1625	{
1626	RT_NOREF1(pUuid);
1627	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",
1628	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1629	int rc;
1630	PQEDIMAGE pImage;
1631
1632	PFNVDPROGRESS pfnProgress = NULL;
1633	void *pvUser = NULL;
1634	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1635	if (pIfProgress)
1636	{
1637	pfnProgress = pIfProgress->pfnProgress;
1638	pvUser = pIfProgress->Core.pvUser;
1639	}
1640
1641	/* Check the VD container type. */
1642	if (enmType != VDTYPE_HDD)
1643	{
1644	rc = VERR_VD_INVALID_TYPE;
1645	goto out;
1646	}
1647
1648	/* Check open flags. All valid flags are supported. */
1649	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1650	{
1651	rc = VERR_INVALID_PARAMETER;
1652	goto out;
1653	}
1654
1655	/* Check remaining arguments. */
1656	if ( !VALID_PTR(pszFilename)
1657	\|\| !*pszFilename
1658	\|\| !VALID_PTR(pPCHSGeometry)
1659	\|\| !VALID_PTR(pLCHSGeometry))
1660	{
1661	rc = VERR_INVALID_PARAMETER;
1662	goto out;
1663	}
1664
1665	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1666	if (!pImage)
1667	{
1668	rc = VERR_NO_MEMORY;
1669	goto out;
1670	}
1671	pImage->pszFilename = pszFilename;
1672	pImage->pStorage = NULL;
1673	pImage->pVDIfsDisk = pVDIfsDisk;
1674	pImage->pVDIfsImage = pVDIfsImage;
1675
1676	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1677	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1678	pfnProgress, pvUser, uPercentStart, uPercentSpan);
1679	if (RT_SUCCESS(rc))
1680	{
1681	/* So far the image is opened in read/write mode. Make sure the
1682	* image is opened in read-only mode if the caller requested that. */
1683	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1684	{
1685	qedFreeImage(pImage, false);
1686	rc = qedOpenImage(pImage, uOpenFlags);
1687	if (RT_FAILURE(rc))
1688	{
1689	RTMemFree(pImage);
1690	goto out;
1691	}
1692	}
1693	*ppBackendData = pImage;
1694	}
1695	else
1696	RTMemFree(pImage);
1697
1698	out:
1699	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1700	return rc;
1701	}
1702
1703	/** @copydoc VBOXHDDBACKEND::pfnRename */
1704	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1705	{
1706	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1707	int rc = VINF_SUCCESS;
1708	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1709
1710	/* Check arguments. */
1711	if ( !pImage
1712	\|\| !pszFilename
1713	\|\| !*pszFilename)
1714	{
1715	rc = VERR_INVALID_PARAMETER;
1716	goto out;
1717	}
1718
1719	/* Close the image. */
1720	rc = qedFreeImage(pImage, false);
1721	if (RT_FAILURE(rc))
1722	goto out;
1723
1724	/* Rename the file. */
1725	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1726	if (RT_FAILURE(rc))
1727	{
1728	/* The move failed, try to reopen the original image. */
1729	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1730	if (RT_FAILURE(rc2))
1731	rc = rc2;
1732
1733	goto out;
1734	}
1735
1736	/* Update pImage with the new information. */
1737	pImage->pszFilename = pszFilename;
1738
1739	/* Open the old image with new name. */
1740	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1741	if (RT_FAILURE(rc))
1742	goto out;
1743
1744	out:
1745	LogFlowFunc(("returns %Rrc\n", rc));
1746	return rc;
1747	}
1748
1749	/** @copydoc VBOXHDDBACKEND::pfnClose */
1750	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1751	{
1752	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1753	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1754	int rc;
1755
1756	rc = qedFreeImage(pImage, fDelete);
1757	RTMemFree(pImage);
1758
1759	LogFlowFunc(("returns %Rrc\n", rc));
1760	return rc;
1761	}
1762
1763	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1764	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1765	{
1766	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1767	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1768	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1769	uint32_t offCluster = 0;
1770	uint32_t idxL1 = 0;
1771	uint32_t idxL2 = 0;
1772	uint64_t offFile = 0;
1773	int rc;
1774
1775	AssertPtr(pImage);
1776	Assert(uOffset % 512 == 0);
1777	Assert(cbToRead % 512 == 0);
1778
1779	if (!VALID_PTR(pIoCtx) \|\| !cbToRead)
1780	{
1781	rc = VERR_INVALID_PARAMETER;
1782	goto out;
1783	}
1784
1785	if ( uOffset + cbToRead > pImage->cbSize
1786	\|\| cbToRead == 0)
1787	{
1788	rc = VERR_INVALID_PARAMETER;
1789	goto out;
1790	}
1791
1792	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1793
1794	/* Clip read size to remain in the cluster. */
1795	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1796
1797	/* Get offset in image. */
1798	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1799	if (RT_SUCCESS(rc))
1800	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1801	pIoCtx, cbToRead);
1802
1803	if ( ( RT_SUCCESS(rc)
1804	\|\| rc == VERR_VD_BLOCK_FREE
1805	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1806	&& pcbActuallyRead)
1807	*pcbActuallyRead = cbToRead;
1808
1809	out:
1810	LogFlowFunc(("returns %Rrc\n", rc));
1811	return rc;
1812	}
1813
1814	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1815	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1816	size_t *pcbPostRead, unsigned fWrite)
1817	{
1818	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1819	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1820	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1821	uint32_t offCluster = 0;
1822	uint32_t idxL1 = 0;
1823	uint32_t idxL2 = 0;
1824	uint64_t offImage = 0;
1825	int rc = VINF_SUCCESS;
1826
1827	AssertPtr(pImage);
1828	Assert(!(uOffset % 512));
1829	Assert(!(cbToWrite % 512));
1830
1831	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
1832	{
1833	rc = VERR_VD_IMAGE_READ_ONLY;
1834	goto out;
1835	}
1836
1837	if (!VALID_PTR(pIoCtx) \|\| !cbToWrite)
1838	{
1839	rc = VERR_INVALID_PARAMETER;
1840	goto out;
1841	}
1842
1843	if ( uOffset + cbToWrite > pImage->cbSize
1844	\|\| cbToWrite == 0)
1845	{
1846	rc = VERR_INVALID_PARAMETER;
1847	goto out;
1848	}
1849
1850	/* Convert offset to L1, L2 index and cluster offset. */
1851	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1852
1853	/* Clip write size to remain in the cluster. */
1854	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1855	Assert(!(cbToWrite % 512));
1856
1857	/* Get offset in image. */
1858	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1859	if (RT_SUCCESS(rc))
1860	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1861	offImage, pIoCtx, cbToWrite, NULL, NULL);
1862	else if (rc == VERR_VD_BLOCK_FREE)
1863	{
1864	if ( cbToWrite == pImage->cbCluster
1865	&& !(fWrite & VD_WRITE_NO_ALLOC))
1866	{
1867	PQEDL2CACHEENTRY pL2Entry = NULL;
1868
1869	/* Full cluster write to previously unallocated cluster.
1870	* Allocate cluster and write data. */
1871	Assert(!offCluster);
1872
1873	do
1874	{
1875	/* Check if we have to allocate a new cluster for L2 tables. */
1876	if (!pImage->paL1Table[idxL1])
1877	{
1878	uint64_t offL2Tbl;
1879	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1880
1881	/* Allocate new async cluster allocation state. */
1882	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1883	if (RT_UNLIKELY(!pL2ClusterAlloc))
1884	{
1885	rc = VERR_NO_MEMORY;
1886	break;
1887	}
1888
1889	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1890	if (!pL2Entry)
1891	{
1892	rc = VERR_NO_MEMORY;
1893	RTMemFree(pL2ClusterAlloc);
1894	break;
1895	}
1896
1897	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1898	pL2Entry->offL2Tbl = offL2Tbl;
1899	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1900
1901	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1902	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1903	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1904	pL2ClusterAlloc->idxL1 = idxL1;
1905	pL2ClusterAlloc->idxL2 = idxL2;
1906	pL2ClusterAlloc->cbToWrite = cbToWrite;
1907	pL2ClusterAlloc->pL2Entry = pL2Entry;
1908
1909	/*
1910	* Write the L2 table first and link to the L1 table afterwards.
1911	* If something unexpected happens the worst case which can happen
1912	* is a leak of some clusters.
1913	*/
1914	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1915	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1916	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1917	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1918	break;
1919	else if (RT_FAILURE(rc))
1920	{
1921	RTMemFree(pL2ClusterAlloc);
1922	qedL2TblCacheEntryFree(pImage, pL2Entry);
1923	break;
1924	}
1925
1926	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1927	}
1928	else
1929	{
1930	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1931	&pL2Entry);
1932
1933	if (RT_SUCCESS(rc))
1934	{
1935	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1936
1937	/* Allocate new async cluster allocation state. */
1938	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1939	if (RT_UNLIKELY(!pDataClusterAlloc))
1940	{
1941	rc = VERR_NO_MEMORY;
1942	break;
1943	}
1944
1945	/* Allocate new cluster for the data. */
1946	uint64_t offData = qedClusterAllocate(pImage, 1);
1947
1948	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1949	pDataClusterAlloc->cbImageOld = offData;
1950	pDataClusterAlloc->offClusterNew = offData;
1951	pDataClusterAlloc->idxL1 = idxL1;
1952	pDataClusterAlloc->idxL2 = idxL2;
1953	pDataClusterAlloc->cbToWrite = cbToWrite;
1954	pDataClusterAlloc->pL2Entry = pL2Entry;
1955
1956	/* Write data. */
1957	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1958	offData, pIoCtx, cbToWrite,
1959	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1960	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1961	break;
1962	else if (RT_FAILURE(rc))
1963	{
1964	RTMemFree(pDataClusterAlloc);
1965	break;
1966	}
1967
1968	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1969	}
1970	}
1971
1972	} while (0);
1973
1974	*pcbPreRead = 0;
1975	*pcbPostRead = 0;
1976	}
1977	else
1978	{
1979	/* Trying to do a partial write to an unallocated cluster. Don't do
1980	* anything except letting the upper layer know what to do. */
1981	*pcbPreRead = offCluster;
1982	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
1983	}
1984	}
1985
1986	if (pcbWriteProcess)
1987	*pcbWriteProcess = cbToWrite;
1988
1989
1990	out:
1991	LogFlowFunc(("returns %Rrc\n", rc));
1992	return rc;
1993	}
1994
1995	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
1996	{
1997	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1998	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1999	int rc = VINF_SUCCESS;
2000
2001	Assert(pImage);
2002
2003	if (VALID_PTR(pIoCtx))
2004	rc = qedFlushImageAsync(pImage, pIoCtx);
2005	else
2006	rc = VERR_INVALID_PARAMETER;
2007
2008	LogFlowFunc(("returns %Rrc\n", rc));
2009	return rc;
2010	}
2011
2012	/** @copydoc VBOXHDDBACKEND::pfnGetVersion */
2013	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
2014	{
2015	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2016	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2017
2018	AssertPtr(pImage);
2019
2020	if (pImage)
2021	return 1;
2022	else
2023	return 0;
2024	}
2025
2026	/** @copydoc VBOXHDDBACKEND::pfnGetSectorSize */
2027	static DECLCALLBACK(uint32_t) qedGetSectorSize(void *pBackendData)
2028	{
2029	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2030	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2031	uint32_t cb = 0;
2032
2033	AssertPtr(pImage);
2034
2035	if (pImage && pImage->pStorage)
2036	cb = 512;
2037
2038	LogFlowFunc(("returns %u\n", cb));
2039	return cb;
2040	}
2041
2042	/** @copydoc VBOXHDDBACKEND::pfnGetSize */
2043	static DECLCALLBACK(uint64_t) qedGetSize(void *pBackendData)
2044	{
2045	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2046	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2047	uint64_t cb = 0;
2048
2049	AssertPtr(pImage);
2050
2051	if (pImage && pImage->pStorage)
2052	cb = pImage->cbSize;
2053
2054	LogFlowFunc(("returns %llu\n", cb));
2055	return cb;
2056	}
2057
2058	/** @copydoc VBOXHDDBACKEND::pfnGetFileSize */
2059	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
2060	{
2061	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2062	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2063	uint64_t cb = 0;
2064
2065	AssertPtr(pImage);
2066
2067	if (pImage)
2068	{
2069	uint64_t cbFile;
2070	if (pImage->pStorage)
2071	{
2072	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
2073	if (RT_SUCCESS(rc))
2074	cb += cbFile;
2075	}
2076	}
2077
2078	LogFlowFunc(("returns %lld\n", cb));
2079	return cb;
2080	}
2081
2082	/** @copydoc VBOXHDDBACKEND::pfnGetPCHSGeometry */
2083	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
2084	PVDGEOMETRY pPCHSGeometry)
2085	{
2086	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
2087	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2088	int rc;
2089
2090	AssertPtr(pImage);
2091
2092	if (pImage)
2093	{
2094	if (pImage->PCHSGeometry.cCylinders)
2095	{
2096	*pPCHSGeometry = pImage->PCHSGeometry;
2097	rc = VINF_SUCCESS;
2098	}
2099	else
2100	rc = VERR_VD_GEOMETRY_NOT_SET;
2101	}
2102	else
2103	rc = VERR_VD_NOT_OPENED;
2104
2105	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2106	return rc;
2107	}
2108
2109	/** @copydoc VBOXHDDBACKEND::pfnSetPCHSGeometry */
2110	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
2111	PCVDGEOMETRY pPCHSGeometry)
2112	{
2113	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n", pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2114	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2115	int rc;
2116
2117	AssertPtr(pImage);
2118
2119	if (pImage)
2120	{
2121	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2122	{
2123	rc = VERR_VD_IMAGE_READ_ONLY;
2124	goto out;
2125	}
2126
2127	pImage->PCHSGeometry = *pPCHSGeometry;
2128	rc = VINF_SUCCESS;
2129	}
2130	else
2131	rc = VERR_VD_NOT_OPENED;
2132
2133	out:
2134	LogFlowFunc(("returns %Rrc\n", rc));
2135	return rc;
2136	}
2137
2138	/** @copydoc VBOXHDDBACKEND::pfnGetLCHSGeometry */
2139	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData,
2140	PVDGEOMETRY pLCHSGeometry)
2141	{
2142	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2143	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2144	int rc;
2145
2146	AssertPtr(pImage);
2147
2148	if (pImage)
2149	{
2150	if (pImage->LCHSGeometry.cCylinders)
2151	{
2152	*pLCHSGeometry = pImage->LCHSGeometry;
2153	rc = VINF_SUCCESS;
2154	}
2155	else
2156	rc = VERR_VD_GEOMETRY_NOT_SET;
2157	}
2158	else
2159	rc = VERR_VD_NOT_OPENED;
2160
2161	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2162	return rc;
2163	}
2164
2165	/** @copydoc VBOXHDDBACKEND::pfnSetLCHSGeometry */
2166	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData,
2167	PCVDGEOMETRY pLCHSGeometry)
2168	{
2169	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData, pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2170	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2171	int rc;
2172
2173	AssertPtr(pImage);
2174
2175	if (pImage)
2176	{
2177	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2178	{
2179	rc = VERR_VD_IMAGE_READ_ONLY;
2180	goto out;
2181	}
2182
2183	pImage->LCHSGeometry = *pLCHSGeometry;
2184	rc = VINF_SUCCESS;
2185	}
2186	else
2187	rc = VERR_VD_NOT_OPENED;
2188
2189	out:
2190	LogFlowFunc(("returns %Rrc\n", rc));
2191	return rc;
2192	}
2193
2194	/** @copydoc VBOXHDDBACKEND::pfnGetImageFlags */
2195	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2196	{
2197	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2198	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2199	unsigned uImageFlags;
2200
2201	AssertPtr(pImage);
2202
2203	if (pImage)
2204	uImageFlags = pImage->uImageFlags;
2205	else
2206	uImageFlags = 0;
2207
2208	LogFlowFunc(("returns %#x\n", uImageFlags));
2209	return uImageFlags;
2210	}
2211
2212	/** @copydoc VBOXHDDBACKEND::pfnGetOpenFlags */
2213	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2214	{
2215	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2216	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2217	unsigned uOpenFlags;
2218
2219	AssertPtr(pImage);
2220
2221	if (pImage)
2222	uOpenFlags = pImage->uOpenFlags;
2223	else
2224	uOpenFlags = 0;
2225
2226	LogFlowFunc(("returns %#x\n", uOpenFlags));
2227	return uOpenFlags;
2228	}
2229
2230	/** @copydoc VBOXHDDBACKEND::pfnSetOpenFlags */
2231	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2232	{
2233	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2234	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2235	int rc;
2236
2237	/* Image must be opened and the new flags must be valid. */
2238	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2239	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2240	{
2241	rc = VERR_INVALID_PARAMETER;
2242	goto out;
2243	}
2244
2245	/* Implement this operation via reopening the image. */
2246	rc = qedFreeImage(pImage, false);
2247	if (RT_FAILURE(rc))
2248	goto out;
2249	rc = qedOpenImage(pImage, uOpenFlags);
2250
2251	out:
2252	LogFlowFunc(("returns %Rrc\n", rc));
2253	return rc;
2254	}
2255
2256	/** @copydoc VBOXHDDBACKEND::pfnGetComment */
2257	static DECLCALLBACK(int) qedGetComment(void pBackendData, char pszComment,
2258	size_t cbComment)
2259	{
2260	RT_NOREF2(pszComment, cbComment);
2261	LogFlowFunc(("pBackendData=%#p pszComment=%#p cbComment=%zu\n", pBackendData, pszComment, cbComment));
2262	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2263	int rc;
2264
2265	AssertPtr(pImage);
2266
2267	if (pImage)
2268	rc = VERR_NOT_SUPPORTED;
2269	else
2270	rc = VERR_VD_NOT_OPENED;
2271
2272	LogFlowFunc(("returns %Rrc comment='%s'\n", rc, pszComment));
2273	return rc;
2274	}
2275
2276	/** @copydoc VBOXHDDBACKEND::pfnSetComment */
2277	static DECLCALLBACK(int) qedSetComment(void pBackendData, const char pszComment)
2278	{
2279	RT_NOREF1(pszComment);
2280	LogFlowFunc(("pBackendData=%#p pszComment=\"%s\"\n", pBackendData, pszComment));
2281	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2282	int rc;
2283
2284	AssertPtr(pImage);
2285
2286	if (pImage)
2287	{
2288	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2289	rc = VERR_VD_IMAGE_READ_ONLY;
2290	else
2291	rc = VERR_NOT_SUPPORTED;
2292	}
2293	else
2294	rc = VERR_VD_NOT_OPENED;
2295
2296	LogFlowFunc(("returns %Rrc\n", rc));
2297	return rc;
2298	}
2299
2300	/** @copydoc VBOXHDDBACKEND::pfnGetUuid */
2301	static DECLCALLBACK(int) qedGetUuid(void *pBackendData, PRTUUID pUuid)
2302	{
2303	RT_NOREF1(pUuid);
2304	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2305	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2306	int rc;
2307
2308	AssertPtr(pImage);
2309
2310	if (pImage)
2311	rc = VERR_NOT_SUPPORTED;
2312	else
2313	rc = VERR_VD_NOT_OPENED;
2314
2315	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2316	return rc;
2317	}
2318
2319	/** @copydoc VBOXHDDBACKEND::pfnSetUuid */
2320	static DECLCALLBACK(int) qedSetUuid(void *pBackendData, PCRTUUID pUuid)
2321	{
2322	RT_NOREF1(pUuid);
2323	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2324	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2325	int rc;
2326
2327	LogFlowFunc(("%RTuuid\n", pUuid));
2328	AssertPtr(pImage);
2329
2330	if (pImage)
2331	{
2332	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2333	rc = VERR_NOT_SUPPORTED;
2334	else
2335	rc = VERR_VD_IMAGE_READ_ONLY;
2336	}
2337	else
2338	rc = VERR_VD_NOT_OPENED;
2339
2340	LogFlowFunc(("returns %Rrc\n", rc));
2341	return rc;
2342	}
2343
2344	/** @copydoc VBOXHDDBACKEND::pfnGetModificationUuid */
2345	static DECLCALLBACK(int) qedGetModificationUuid(void *pBackendData, PRTUUID pUuid)
2346	{
2347	RT_NOREF1(pUuid);
2348	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2349	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2350	int rc;
2351
2352	AssertPtr(pImage);
2353
2354	if (pImage)
2355	rc = VERR_NOT_SUPPORTED;
2356	else
2357	rc = VERR_VD_NOT_OPENED;
2358
2359	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2360	return rc;
2361	}
2362
2363	/** @copydoc VBOXHDDBACKEND::pfnSetModificationUuid */
2364	static DECLCALLBACK(int) qedSetModificationUuid(void *pBackendData, PCRTUUID pUuid)
2365	{
2366	RT_NOREF1(pUuid);
2367	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2368	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2369	int rc;
2370
2371	AssertPtr(pImage);
2372
2373	if (pImage)
2374	{
2375	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2376	rc = VERR_NOT_SUPPORTED;
2377	else
2378	rc = VERR_VD_IMAGE_READ_ONLY;
2379	}
2380	else
2381	rc = VERR_VD_NOT_OPENED;
2382
2383	LogFlowFunc(("returns %Rrc\n", rc));
2384	return rc;
2385	}
2386
2387	/** @copydoc VBOXHDDBACKEND::pfnGetParentUuid */
2388	static DECLCALLBACK(int) qedGetParentUuid(void *pBackendData, PRTUUID pUuid)
2389	{
2390	RT_NOREF1(pUuid);
2391	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2392	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2393	int rc;
2394
2395	AssertPtr(pImage);
2396
2397	if (pImage)
2398	rc = VERR_NOT_SUPPORTED;
2399	else
2400	rc = VERR_VD_NOT_OPENED;
2401
2402	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2403	return rc;
2404	}
2405
2406	/** @copydoc VBOXHDDBACKEND::pfnSetParentUuid */
2407	static DECLCALLBACK(int) qedSetParentUuid(void *pBackendData, PCRTUUID pUuid)
2408	{
2409	RT_NOREF1(pUuid);
2410	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2411	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2412	int rc;
2413
2414	AssertPtr(pImage);
2415
2416	if (pImage)
2417	{
2418	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2419	rc = VERR_NOT_SUPPORTED;
2420	else
2421	rc = VERR_VD_IMAGE_READ_ONLY;
2422	}
2423	else
2424	rc = VERR_VD_NOT_OPENED;
2425
2426	LogFlowFunc(("returns %Rrc\n", rc));
2427	return rc;
2428	}
2429
2430	/** @copydoc VBOXHDDBACKEND::pfnGetParentModificationUuid */
2431	static DECLCALLBACK(int) qedGetParentModificationUuid(void *pBackendData, PRTUUID pUuid)
2432	{
2433	RT_NOREF1(pUuid);
2434	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2435	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2436	int rc;
2437
2438	AssertPtr(pImage);
2439
2440	if (pImage)
2441	rc = VERR_NOT_SUPPORTED;
2442	else
2443	rc = VERR_VD_NOT_OPENED;
2444
2445	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2446	return rc;
2447	}
2448
2449	/** @copydoc VBOXHDDBACKEND::pfnSetParentModificationUuid */
2450	static DECLCALLBACK(int) qedSetParentModificationUuid(void *pBackendData, PCRTUUID pUuid)
2451	{
2452	RT_NOREF1(pUuid);
2453	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2454	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2455	int rc;
2456
2457	AssertPtr(pImage);
2458
2459	if (pImage)
2460	{
2461	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2462	rc = VERR_NOT_SUPPORTED;
2463	else
2464	rc = VERR_VD_IMAGE_READ_ONLY;
2465	}
2466	else
2467	rc = VERR_VD_NOT_OPENED;
2468
2469	LogFlowFunc(("returns %Rrc\n", rc));
2470	return rc;
2471	}
2472
2473	/** @copydoc VBOXHDDBACKEND::pfnDump */
2474	static DECLCALLBACK(void) qedDump(void *pBackendData)
2475	{
2476	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2477
2478	AssertPtr(pImage);
2479	if (pImage)
2480	{
2481	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2482	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2483	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2484	pImage->cbSize / 512);
2485	}
2486	}
2487
2488	/** @copydoc VBOXHDDBACKEND::pfnGetParentFilename */
2489	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2490	{
2491	int rc = VINF_SUCCESS;
2492	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2493
2494	AssertPtr(pImage);
2495	if (pImage)
2496	if (pImage->pszBackingFilename)
2497	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2498	else
2499	rc = VERR_NOT_SUPPORTED;
2500	else
2501	rc = VERR_VD_NOT_OPENED;
2502
2503	LogFlowFunc(("returns %Rrc\n", rc));
2504	return rc;
2505	}
2506
2507	/** @copydoc VBOXHDDBACKEND::pfnSetParentFilename */
2508	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2509	{
2510	int rc = VINF_SUCCESS;
2511	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2512
2513	AssertPtr(pImage);
2514	if (pImage)
2515	{
2516	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2517	rc = VERR_VD_IMAGE_READ_ONLY;
2518	else if ( pImage->pszBackingFilename
2519	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2520	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2521	else
2522	{
2523	if (pImage->pszBackingFilename)
2524	RTStrFree(pImage->pszBackingFilename);
2525	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2526	if (!pImage->pszBackingFilename)
2527	rc = VERR_NO_MEMORY;
2528	else
2529	{
2530	if (!pImage->offBackingFilename)
2531	{
2532	/* Allocate new cluster. */
2533	uint64_t offData = qedClusterAllocate(pImage, 1);
2534
2535	Assert((offData & UINT32_MAX) == offData);
2536	pImage->offBackingFilename = (uint32_t)offData;
2537	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2538	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2539	offData + pImage->cbCluster);
2540	}
2541
2542	if (RT_SUCCESS(rc))
2543	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2544	pImage->offBackingFilename,
2545	pImage->pszBackingFilename,
2546	strlen(pImage->pszBackingFilename));
2547	}
2548	}
2549	}
2550	else
2551	rc = VERR_VD_NOT_OPENED;
2552
2553	LogFlowFunc(("returns %Rrc\n", rc));
2554	return rc;
2555	}
2556
2557	/** @copydoc VBOXHDDBACKEND::pfnResize */
2558	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2559	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2560	unsigned uPercentStart, unsigned uPercentSpan,
2561	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2562	PVDINTERFACE pVDIfsOperation)
2563	{
2564	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2565	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2566	int rc = VINF_SUCCESS;
2567
2568	/* Making the image smaller is not supported at the moment. */
2569	if (cbSize < pImage->cbSize)
2570	rc = VERR_NOT_SUPPORTED;
2571	else if (cbSize > pImage->cbSize)
2572	{
2573	/*
2574	* It is enough to just update the size field in the header to complete
2575	* growing. With the default cluster and table sizes the image can be expanded
2576	* to 64TB without overflowing the L1 and L2 tables making block relocation
2577	* superfluous.
2578	* @todo: The rare case where block relocation is still required (non default
2579	* table and/or cluster size or images with more than 64TB) is not
2580	* implemented yet and resizing such an image will fail with an error.
2581	*/
2582	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2583	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2584	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2585	pImage->pszFilename);
2586	else
2587	{
2588	uint64_t cbSizeOld = pImage->cbSize;
2589
2590	pImage->cbSize = cbSize;
2591	rc = qedFlushImage(pImage);
2592	if (RT_FAILURE(rc))
2593	{
2594	pImage->cbSize = cbSizeOld; /* Restore */
2595
2596	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2597	pImage->pszFilename);
2598	}
2599	}
2600	}
2601	/* Same size doesn't change the image at all. */
2602
2603	LogFlowFunc(("returns %Rrc\n", rc));
2604	return rc;
2605	}
2606
2607
2608	const VBOXHDDBACKEND g_QedBackend =
2609	{
2610	/* pszBackendName */
2611	"QED",
2612	/* cbSize */
2613	sizeof(VBOXHDDBACKEND),
2614	/* uBackendCaps */
2615	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2616	/* paFileExtensions */
2617	s_aQedFileExtensions,
2618	/* paConfigInfo */
2619	NULL,
2620	/* pfnCheckIfValid */
2621	qedCheckIfValid,
2622	/* pfnOpen */
2623	qedOpen,
2624	/* pfnCreate */
2625	qedCreate,
2626	/* pfnRename */
2627	qedRename,
2628	/* pfnClose */
2629	qedClose,
2630	/* pfnRead */
2631	qedRead,
2632	/* pfnWrite */
2633	qedWrite,
2634	/* pfnFlush */
2635	qedFlush,
2636	/* pfnDiscard */
2637	NULL,
2638	/* pfnGetVersion */
2639	qedGetVersion,
2640	/* pfnGetSectorSize */
2641	qedGetSectorSize,
2642	/* pfnGetSize */
2643	qedGetSize,
2644	/* pfnGetFileSize */
2645	qedGetFileSize,
2646	/* pfnGetPCHSGeometry */
2647	qedGetPCHSGeometry,
2648	/* pfnSetPCHSGeometry */
2649	qedSetPCHSGeometry,
2650	/* pfnGetLCHSGeometry */
2651	qedGetLCHSGeometry,
2652	/* pfnSetLCHSGeometry */
2653	qedSetLCHSGeometry,
2654	/* pfnGetImageFlags */
2655	qedGetImageFlags,
2656	/* pfnGetOpenFlags */
2657	qedGetOpenFlags,
2658	/* pfnSetOpenFlags */
2659	qedSetOpenFlags,
2660	/* pfnGetComment */
2661	qedGetComment,
2662	/* pfnSetComment */
2663	qedSetComment,
2664	/* pfnGetUuid */
2665	qedGetUuid,
2666	/* pfnSetUuid */
2667	qedSetUuid,
2668	/* pfnGetModificationUuid */
2669	qedGetModificationUuid,
2670	/* pfnSetModificationUuid */
2671	qedSetModificationUuid,
2672	/* pfnGetParentUuid */
2673	qedGetParentUuid,
2674	/* pfnSetParentUuid */
2675	qedSetParentUuid,
2676	/* pfnGetParentModificationUuid */
2677	qedGetParentModificationUuid,
2678	/* pfnSetParentModificationUuid */
2679	qedSetParentModificationUuid,
2680	/* pfnDump */
2681	qedDump,
2682	/* pfnGetTimestamp */
2683	NULL,
2684	/* pfnGetParentTimestamp */
2685	NULL,
2686	/* pfnSetParentTimestamp */
2687	NULL,
2688	/* pfnGetParentFilename */
2689	qedGetParentFilename,
2690	/* pfnSetParentFilename */
2691	qedSetParentFilename,
2692	/* pfnComposeLocation */
2693	genericFileComposeLocation,
2694	/* pfnComposeName */
2695	genericFileComposeName,
2696	/* pfnCompact */
2697	NULL,
2698	/* pfnResize */
2699	qedResize,
2700	/* pfnRepair */
2701	NULL,
2702	/* pfnTraverseMetadata */
2703	NULL
2704	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 63768

Download in other formats: