QED.cpp@ 63699

Last change on this file since 63699 was 63636, checked in by vboxsync, 8 years ago
Storage/QED: Check that the filename size for the backing file is in bounds if the feature flag is set
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 90.2 KB

Line
1	/* $Id: QED.cpp 63636 2016-08-25 14:09:21Z 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	#if 0 /* unused */
569	/**
570	* Fetches the L2 from the given offset trying the LRU cache first and
571	* reading it from the image after a cache miss.
572	*
573	* @returns VBox status code.
574	* @param pImage Image instance data.
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 qedL2TblCacheFetch(PQEDIMAGE pImage, uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
579	{
580	int rc = VINF_SUCCESS;
581
582	LogFlowFunc(("pImage=%#p offL2Tbl=%llu ppL2Entry=%#p\n", pImage, offL2Tbl, ppL2Entry));
583
584	/* Try to fetch the L2 table from the cache first. */
585	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
586	if (!pL2Entry)
587	{
588	LogFlowFunc(("Reading L2 table from image\n"));
589	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
590
591	if (pL2Entry)
592	{
593	/* Read from the image. */
594	pL2Entry->offL2Tbl = offL2Tbl;
595	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, offL2Tbl,
596	pL2Entry->paL2Tbl, pImage->cbTable);
597	if (RT_SUCCESS(rc))
598	{
599	#if defined(RT_BIG_ENDIAN)
600	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
601	#endif
602	qedL2TblCacheEntryInsert(pImage, pL2Entry);
603	}
604	else
605	{
606	qedL2TblCacheEntryRelease(pL2Entry);
607	qedL2TblCacheEntryFree(pImage, pL2Entry);
608	}
609	}
610	else
611	rc = VERR_NO_MEMORY;
612	}
613
614	if (RT_SUCCESS(rc))
615	*ppL2Entry = pL2Entry;
616
617	LogFlowFunc(("returns rc=%Rrc\n", rc));
618	return rc;
619	}
620	#endif
621
622	/**
623	* Fetches the L2 from the given offset trying the LRU cache first and
624	* reading it from the image after a cache miss - version for async I/O.
625	*
626	* @returns VBox status code.
627	* @param pImage Image instance data.
628	* @param pIoCtx The I/O context.
629	* @param offL2Tbl The offset of the L2 table in the image.
630	* @param ppL2Entry Where to store the L2 table on success.
631	*/
632	static int qedL2TblCacheFetchAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
633	uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
634	{
635	int rc = VINF_SUCCESS;
636
637	/* Try to fetch the L2 table from the cache first. */
638	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
639	if (!pL2Entry)
640	{
641	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
642
643	if (pL2Entry)
644	{
645	/* Read from the image. */
646	PVDMETAXFER pMetaXfer;
647
648	pL2Entry->offL2Tbl = offL2Tbl;
649	rc = vdIfIoIntFileReadMeta(pImage->pIfIo, pImage->pStorage,
650	offL2Tbl, pL2Entry->paL2Tbl,
651	pImage->cbTable, pIoCtx,
652	&pMetaXfer, NULL, NULL);
653	if (RT_SUCCESS(rc))
654	{
655	vdIfIoIntMetaXferRelease(pImage->pIfIo, pMetaXfer);
656	#if defined(RT_BIG_ENDIAN)
657	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
658	#endif
659	qedL2TblCacheEntryInsert(pImage, pL2Entry);
660	}
661	else
662	{
663	qedL2TblCacheEntryRelease(pL2Entry);
664	qedL2TblCacheEntryFree(pImage, pL2Entry);
665	}
666	}
667	else
668	rc = VERR_NO_MEMORY;
669	}
670
671	if (RT_SUCCESS(rc))
672	*ppL2Entry = pL2Entry;
673
674	return rc;
675	}
676
677	/**
678	* Return power of 2 or 0 if num error.
679	*
680	* @returns The power of 2 or 0 if the given number is not a power of 2.
681	* @param u32 The number.
682	*/
683	static uint32_t qedGetPowerOfTwo(uint32_t u32)
684	{
685	if (u32 == 0)
686	return 0;
687	uint32_t uPower2 = 0;
688	while ((u32 & 1) == 0)
689	{
690	u32 >>= 1;
691	uPower2++;
692	}
693	return u32 == 1 ? uPower2 : 0;
694	}
695
696	/**
697	* Sets the L1, L2 and offset bitmasks and L1 and L2 bit shift members.
698	*
699	* @returns nothing.
700	* @param pImage The image instance data.
701	*/
702	static void qedTableMasksInit(PQEDIMAGE pImage)
703	{
704	uint32_t cClusterBits, cTableBits;
705
706	cClusterBits = qedGetPowerOfTwo(pImage->cbCluster);
707	cTableBits = qedGetPowerOfTwo(pImage->cTableEntries);
708
709	Assert(cClusterBits + 2 * cTableBits <= 64);
710
711	pImage->fOffsetMask = ((uint64_t)pImage->cbCluster - 1);
712	pImage->fL2Mask = ((uint64_t)pImage->cTableEntries - 1) << cClusterBits;
713	pImage->cL2Shift = cClusterBits;
714	pImage->fL1Mask = ((uint64_t)pImage->cTableEntries - 1) << (cClusterBits + cTableBits);
715	pImage->cL1Shift = cClusterBits + cTableBits;
716	}
717
718	/**
719	* Converts a given logical offset into the
720	*
721	* @returns nothing.
722	* @param pImage The image instance data.
723	* @param off The logical offset to convert.
724	* @param pidxL1 Where to store the index in the L1 table on success.
725	* @param pidxL2 Where to store the index in the L2 table on success.
726	* @param poffCluster Where to store the offset in the cluster on success.
727	*/
728	DECLINLINE(void) qedConvertLogicalOffset(PQEDIMAGE pImage, uint64_t off, uint32_t *pidxL1,
729	uint32_t pidxL2, uint32_t poffCluster)
730	{
731	AssertPtr(pidxL1);
732	AssertPtr(pidxL2);
733	AssertPtr(poffCluster);
734
735	*poffCluster = off & pImage->fOffsetMask;
736	*pidxL1 = (off & pImage->fL1Mask) >> pImage->cL1Shift;
737	*pidxL2 = (off & pImage->fL2Mask) >> pImage->cL2Shift;
738	}
739
740	/**
741	* Converts Cluster size to a byte size.
742	*
743	* @returns Number of bytes derived from the given number of clusters.
744	* @param pImage The image instance data.
745	* @param cClusters The clusters to convert.
746	*/
747	DECLINLINE(uint64_t) qedCluster2Byte(PQEDIMAGE pImage, uint64_t cClusters)
748	{
749	return cClusters * pImage->cbCluster;
750	}
751
752	/**
753	* Converts number of bytes to cluster size rounding to the next cluster.
754	*
755	* @returns Number of bytes derived from the given number of clusters.
756	* @param pImage The image instance data.
757	* @param cb Number of bytes to convert.
758	*/
759	DECLINLINE(uint64_t) qedByte2Cluster(PQEDIMAGE pImage, uint64_t cb)
760	{
761	return cb / pImage->cbCluster + (cb % pImage->cbCluster ? 1 : 0);
762	}
763
764	/**
765	* Allocates a new cluster in the image.
766	*
767	* @returns The start offset of the new cluster in the image.
768	* @param pImage The image instance data.
769	* @param cCLusters Number of clusters to allocate.
770	*/
771	DECLINLINE(uint64_t) qedClusterAllocate(PQEDIMAGE pImage, uint32_t cClusters)
772	{
773	uint64_t offCluster;
774
775	offCluster = pImage->cbImage;
776	pImage->cbImage += cClusters*pImage->cbCluster;
777
778	return offCluster;
779	}
780
781	/**
782	* Returns the real image offset for a given cluster or an error if the cluster is not
783	* yet allocated.
784	*
785	* @returns VBox status code.
786	* VERR_VD_BLOCK_FREE if the cluster is not yet allocated.
787	* @param pImage The image instance data.
788	* @param pIoCtx The I/O context.
789	* @param idxL1 The L1 index.
790	* @param idxL2 The L2 index.
791	* @param offCluster Offset inside the cluster.
792	* @param poffImage Where to store the image offset on success;
793	*/
794	static int qedConvertToImageOffset(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
795	uint32_t idxL1, uint32_t idxL2,
796	uint32_t offCluster, uint64_t *poffImage)
797	{
798	int rc = VERR_VD_BLOCK_FREE;
799
800	AssertReturn(idxL1 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
801	AssertReturn(idxL2 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
802
803	if (pImage->paL1Table[idxL1])
804	{
805	PQEDL2CACHEENTRY pL2Entry;
806
807	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
808	&pL2Entry);
809	if (RT_SUCCESS(rc))
810	{
811	/* Get real file offset. */
812	if (pL2Entry->paL2Tbl[idxL2])
813	*poffImage = pL2Entry->paL2Tbl[idxL2] + offCluster;
814	else
815	rc = VERR_VD_BLOCK_FREE;
816
817	qedL2TblCacheEntryRelease(pL2Entry);
818	}
819	}
820
821	return rc;
822	}
823
824
825	/**
826	* Internal. Flush image data to disk.
827	*/
828	static int qedFlushImage(PQEDIMAGE pImage)
829	{
830	int rc = VINF_SUCCESS;
831
832	if ( pImage->pStorage
833	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
834	{
835	QedHeader Header;
836
837	Assert(!(pImage->cbTable % pImage->cbCluster));
838	#if defined(RT_BIG_ENDIAN)
839	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
840	if (paL1TblImg)
841	{
842	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
843	pImage->cTableEntries);
844	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
845	pImage->offL1Table, paL1TblImg,
846	pImage->cbTable);
847	RTMemFree(paL1TblImg);
848	}
849	else
850	rc = VERR_NO_MEMORY;
851	#else
852	/* Write L1 table directly. */
853	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, pImage->offL1Table,
854	pImage->paL1Table, pImage->cbTable);
855	#endif
856	if (RT_SUCCESS(rc))
857	{
858	/* Write header. */
859	qedHdrConvertFromHostEndianess(pImage, &Header);
860	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, 0, &Header,
861	sizeof(Header));
862	if (RT_SUCCESS(rc))
863	rc = vdIfIoIntFileFlushSync(pImage->pIfIo, pImage->pStorage);
864	}
865	}
866
867	return rc;
868	}
869
870	/**
871	* Flush image data to disk - version for async I/O.
872	*
873	* @returns VBox status code.
874	* @param pImage The image instance data.
875	* @param pIoCtx The I/o context
876	*/
877	static int qedFlushImageAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx)
878	{
879	int rc = VINF_SUCCESS;
880
881	if ( pImage->pStorage
882	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
883	{
884	QedHeader Header;
885
886	Assert(!(pImage->cbTable % pImage->cbCluster));
887	#if defined(RT_BIG_ENDIAN)
888	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
889	if (paL1TblImg)
890	{
891	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
892	pImage->cTableEntries);
893	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
894	pImage->offL1Table, paL1TblImg,
895	pImage->cbTable, pIoCtx, NULL, NULL);
896	RTMemFree(paL1TblImg);
897	}
898	else
899	rc = VERR_NO_MEMORY;
900	#else
901	/* Write L1 table directly. */
902	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
903	pImage->offL1Table, pImage->paL1Table,
904	pImage->cbTable, pIoCtx, NULL, NULL);
905	#endif
906	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
907	{
908	/* Write header. */
909	qedHdrConvertFromHostEndianess(pImage, &Header);
910	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
911	0, &Header, sizeof(Header),
912	pIoCtx, NULL, NULL);
913	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
914	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
915	pIoCtx, NULL, NULL);
916	}
917	}
918
919	return rc;
920	}
921
922	/**
923	* Checks whether the given cluster offset is valid.
924	*
925	* @returns Whether the given cluster offset is valid.
926	* @param offCluster The table offset to check.
927	* @param cbFile The real file size of the image.
928	* @param cbCluster The cluster size in bytes.
929	*/
930	DECLINLINE(bool) qedIsClusterOffsetValid(uint64_t offCluster, uint64_t cbFile, size_t cbCluster)
931	{
932	return (offCluster <= cbFile - cbCluster)
933	&& !(offCluster & (cbCluster - 1));
934	}
935
936	/**
937	* Checks whether the given table offset is valid.
938	*
939	* @returns Whether the given table offset is valid.
940	* @param offTbl The table offset to check.
941	* @param cbFile The real file size of the image.
942	* @param cbTable The table size in bytes.
943	* @param cbCluster The cluster size in bytes.
944	*/
945	DECLINLINE(bool) qedIsTblOffsetValid(uint64_t offTbl, uint64_t cbFile, size_t cbTable, size_t cbCluster)
946	{
947	return (offTbl <= cbFile - cbTable)
948	&& !(offTbl & (cbCluster - 1));
949	}
950
951	/**
952	* Sets the specified range in the cluster bitmap checking whether any of the clusters is already
953	* used before.
954	*
955	* @returns Whether the range was clear and is set now.
956	* @param pvClusterBitmap The cluster bitmap to use.
957	* @param offClusterStart The first cluster to check and set.
958	* @param offClusterEnd The first cluster to not check and set anymore.
959	*/
960	static bool qedClusterBitmapCheckAndSet(void *pvClusterBitmap, uint32_t offClusterStart, uint32_t offClusterEnd)
961	{
962	for (uint32_t offCluster = offClusterStart; offCluster < offClusterEnd; offCluster++)
963	if (ASMBitTest(pvClusterBitmap, offCluster))
964	return false;
965
966	ASMBitSetRange(pvClusterBitmap, offClusterStart, offClusterEnd);
967	return true;
968	}
969
970	/**
971	* Checks the given image for consistency, usually called when the
972	* QED_FEATURE_NEED_CHECK bit is set.
973	*
974	* @returns VBox status code.
975	* @retval VINF_SUCCESS when the image can be accessed.
976	* @param pImage The image instance data.
977	* @param pHeader The header to use for checking.
978	*
979	* @note It is not required that the image state is fully initialized Only
980	* The I/O interface and storage handle need to be valid.
981	* @note The header must be converted to the host CPU endian format already
982	* and should be validated already.
983	*/
984	static int qedCheckImage(PQEDIMAGE pImage, PQedHeader pHeader)
985	{
986	uint64_t cbFile;
987	uint32_t cbTable;
988	uint32_t cTableEntries;
989	uint64_t *paL1Tbl = NULL;
990	uint64_t *paL2Tbl = NULL;
991	void *pvClusterBitmap = NULL;
992	uint32_t offClusterStart;
993	int rc = VINF_SUCCESS;
994
995	pImage->cbCluster = pHeader->u32ClusterSize;
996	cbTable = pHeader->u32TableSize * pHeader->u32ClusterSize;
997	cTableEntries = cbTable / sizeof(uint64_t);
998
999	do
1000	{
1001	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1002	if (RT_FAILURE(rc))
1003	{
1004	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1005	N_("Qed: Querying the file size of image '%s' failed"),
1006	pImage->pszFilename);
1007	break;
1008	}
1009
1010	/* Allocate L1 table. */
1011	paL1Tbl = (uint64_t *)RTMemAllocZ(cbTable);
1012	if (!paL1Tbl)
1013	{
1014	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1015	N_("Qed: Allocating memory for the L1 table for image '%s' failed"),
1016	pImage->pszFilename);
1017	break;
1018	}
1019
1020	paL2Tbl = (uint64_t *)RTMemAllocZ(cbTable);
1021	if (!paL2Tbl)
1022	{
1023	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1024	N_("Qed: Allocating memory for the L2 table for image '%s' failed"),
1025	pImage->pszFilename);
1026	break;
1027	}
1028
1029	pvClusterBitmap = RTMemAllocZ(cbFile / pHeader->u32ClusterSize / 8);
1030	if (!pvClusterBitmap)
1031	{
1032	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1033	N_("Qed: Allocating memory for the cluster bitmap for image '%s' failed"),
1034	pImage->pszFilename);
1035	break;
1036	}
1037
1038	/* Validate L1 table offset. */
1039	if (!qedIsTblOffsetValid(pHeader->u64OffL1Table, cbFile, cbTable, pHeader->u32ClusterSize))
1040	{
1041	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1042	N_("Qed: L1 table offset of image '%s' is corrupt (%llu)"),
1043	pImage->pszFilename, pHeader->u64OffL1Table);
1044	break;
1045	}
1046
1047	/* Read L1 table. */
1048	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1049	pHeader->u64OffL1Table, paL1Tbl, cbTable);
1050	if (RT_FAILURE(rc))
1051	{
1052	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1053	N_("Qed: Reading the L1 table from image '%s' failed"),
1054	pImage->pszFilename);
1055	break;
1056	}
1057
1058	/* Mark the L1 table in cluster bitmap. */
1059	ASMBitSet(pvClusterBitmap, 0); /* Header is always in cluster 0. */
1060	offClusterStart = qedByte2Cluster(pImage, pHeader->u64OffL1Table);
1061	bool fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1062	Assert(fSet);
1063
1064	/* Scan the L1 and L2 tables for invalid entries. */
1065	qedTableConvertToHostEndianess(paL1Tbl, cTableEntries);
1066
1067	for (unsigned iL1 = 0; iL1 < cTableEntries; iL1++)
1068	{
1069	if (!paL1Tbl[iL1])
1070	continue; /* Skip unallocated clusters. */
1071
1072	if (!qedIsTblOffsetValid(paL1Tbl[iL1], cbFile, cbTable, pHeader->u32ClusterSize))
1073	{
1074	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1075	N_("Qed: Entry %d of the L1 table from image '%s' is invalid (%llu)"),
1076	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1077	break;
1078	}
1079
1080	/* Now check that the clusters are not allocated already. */
1081	offClusterStart = qedByte2Cluster(pImage, paL1Tbl[iL1]);
1082	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1083	if (!fSet)
1084	{
1085	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1086	N_("Qed: Entry %d of the L1 table from image '%s' points to a already used cluster (%llu)"),
1087	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1088	break;
1089	}
1090
1091	/* Read the linked L2 table and check it. */
1092	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1093	paL1Tbl[iL1], paL2Tbl, cbTable);
1094	if (RT_FAILURE(rc))
1095	{
1096	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1097	N_("Qed: Reading the L2 table from image '%s' failed"),
1098	pImage->pszFilename);
1099	break;
1100	}
1101
1102	/* Check all L2 entries. */
1103	for (unsigned iL2 = 0; iL2 < cTableEntries; iL2++)
1104	{
1105	if (paL2Tbl[iL2])
1106	continue; /* Skip unallocated clusters. */
1107
1108	if (!qedIsClusterOffsetValid(paL2Tbl[iL2], cbFile, pHeader->u32ClusterSize))
1109	{
1110	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1111	N_("Qed: Entry %d of the L2 table from image '%s' is invalid (%llu)"),
1112	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1113	break;
1114	}
1115
1116	/* Now check that the clusters are not allocated already. */
1117	offClusterStart = qedByte2Cluster(pImage, paL2Tbl[iL2]);
1118	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + 1);
1119	if (!fSet)
1120	{
1121	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1122	N_("Qed: Entry %d of the L2 table from image '%s' points to a already used cluster (%llu)"),
1123	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1124	break;
1125	}
1126	}
1127	}
1128	} while(0);
1129
1130	if (paL1Tbl)
1131	RTMemFree(paL1Tbl);
1132	if (paL2Tbl)
1133	RTMemFree(paL2Tbl);
1134	if (pvClusterBitmap)
1135	RTMemFree(pvClusterBitmap);
1136
1137	return rc;
1138	}
1139
1140	/**
1141	* Internal. Free all allocated space for representing an image except pImage,
1142	* and optionally delete the image from disk.
1143	*/
1144	static int qedFreeImage(PQEDIMAGE pImage, bool fDelete)
1145	{
1146	int rc = VINF_SUCCESS;
1147
1148	/* Freeing a never allocated image (e.g. because the open failed) is
1149	* not signalled as an error. After all nothing bad happens. */
1150	if (pImage)
1151	{
1152	if (pImage->pStorage)
1153	{
1154	/* No point updating the file that is deleted anyway. */
1155	if (!fDelete)
1156	qedFlushImage(pImage);
1157
1158	rc = vdIfIoIntFileClose(pImage->pIfIo, pImage->pStorage);
1159	pImage->pStorage = NULL;
1160	}
1161
1162	if (pImage->paL1Table)
1163	RTMemFree(pImage->paL1Table);
1164
1165	if (pImage->pszBackingFilename)
1166	{
1167	RTMemFree(pImage->pszBackingFilename);
1168	pImage->pszBackingFilename = NULL;
1169	}
1170
1171	qedL2TblCacheDestroy(pImage);
1172
1173	if (fDelete && pImage->pszFilename)
1174	vdIfIoIntFileDelete(pImage->pIfIo, pImage->pszFilename);
1175	}
1176
1177	LogFlowFunc(("returns %Rrc\n", rc));
1178	return rc;
1179	}
1180
1181	/**
1182	* Internal: Open an image, constructing all necessary data structures.
1183	*/
1184	static int qedOpenImage(PQEDIMAGE pImage, unsigned uOpenFlags)
1185	{
1186	int rc;
1187
1188	pImage->uOpenFlags = uOpenFlags;
1189
1190	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1191	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1192	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1193
1194	/*
1195	* Create the L2 cache before opening the image so we can call qedFreeImage()
1196	* even if opening the image file fails.
1197	*/
1198	rc = qedL2TblCacheCreate(pImage);
1199	if (RT_FAILURE(rc))
1200	{
1201	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1202	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1203	pImage->pszFilename);
1204
1205	goto out;
1206	}
1207
1208	/*
1209	* Open the image.
1210	*/
1211	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename,
1212	VDOpenFlagsToFileOpenFlags(uOpenFlags,
1213	false /* fCreate */),
1214	&pImage->pStorage);
1215	if (RT_FAILURE(rc))
1216	{
1217	/* Do NOT signal an appropriate error here, as the VD layer has the
1218	* choice of retrying the open if it failed. */
1219	goto out;
1220	}
1221
1222	uint64_t cbFile;
1223	QedHeader Header;
1224	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1225	if (RT_FAILURE(rc))
1226	goto out;
1227	if (cbFile > sizeof(Header))
1228	{
1229	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, 0, &Header, sizeof(Header));
1230	if ( RT_SUCCESS(rc)
1231	&& qedHdrConvertToHostEndianess(&Header))
1232	{
1233	if ( !(Header.u64FeatureFlags & ~QED_FEATURE_MASK)
1234	&& !(Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE_NO_PROBE))
1235	{
1236	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1237	{
1238	/* Image needs checking. */
1239	if (!(uOpenFlags & VD_OPEN_FLAGS_READONLY))
1240	rc = qedCheckImage(pImage, &Header);
1241	else
1242	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1243	N_("Qed: Image '%s' needs checking but is opened readonly"),
1244	pImage->pszFilename);
1245	}
1246
1247	if ( RT_SUCCESS(rc)
1248	&& (Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE))
1249	{
1250	/* Load backing filename from image. */
1251	pImage->pszBackingFilename = (char )RTMemAllocZ(Header.u32BackingFilenameSize + 1); / +1 for \0 terminator. */
1252	if (pImage->pszBackingFilename)
1253	{
1254	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1255	pImage->offBackingFilename = Header.u32OffBackingFilename;
1256	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1257	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1258	Header.u32BackingFilenameSize);
1259	}
1260	else
1261	rc = VERR_NO_MEMORY;
1262	}
1263
1264	if (RT_SUCCESS(rc))
1265	{
1266	pImage->cbImage = cbFile;
1267	pImage->cbCluster = Header.u32ClusterSize;
1268	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1269	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1270	pImage->offL1Table = Header.u64OffL1Table;
1271	pImage->cbSize = Header.u64Size;
1272	qedTableMasksInit(pImage);
1273
1274	/* Allocate L1 table. */
1275	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1276	if (pImage->paL1Table)
1277	{
1278	/* Read from the image. */
1279	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1280	pImage->offL1Table, pImage->paL1Table,
1281	pImage->cbTable);
1282	if (RT_SUCCESS(rc))
1283	{
1284	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1285
1286	/* If the consistency check succeeded, clear the flag by flushing the image. */
1287	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1288	rc = qedFlushImage(pImage);
1289	}
1290	else
1291	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1292	N_("Qed: Reading the L1 table for image '%s' failed"),
1293	pImage->pszFilename);
1294	}
1295	else
1296	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1297	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1298	pImage->pszFilename);
1299	}
1300	}
1301	else
1302	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1303	N_("Qed: The image '%s' makes use of unsupported features"),
1304	pImage->pszFilename);
1305	}
1306	else if (RT_SUCCESS(rc))
1307	rc = VERR_VD_GEN_INVALID_HEADER;
1308	}
1309	else
1310	rc = VERR_VD_GEN_INVALID_HEADER;
1311
1312	out:
1313	if (RT_FAILURE(rc))
1314	qedFreeImage(pImage, false);
1315	return rc;
1316	}
1317
1318	/**
1319	* Internal: Create a qed image.
1320	*/
1321	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1322	unsigned uImageFlags, const char *pszComment,
1323	PCVDGEOMETRY pPCHSGeometry,
1324	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1325	PFNVDPROGRESS pfnProgress, void *pvUser,
1326	unsigned uPercentStart, unsigned uPercentSpan)
1327	{
1328	RT_NOREF1(pszComment);
1329	int rc;
1330	int32_t fOpen;
1331
1332	if (uImageFlags & VD_IMAGE_FLAGS_FIXED)
1333	{
1334	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1335	goto out;
1336	}
1337
1338	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1339	pImage->uImageFlags = uImageFlags;
1340	pImage->PCHSGeometry = *pPCHSGeometry;
1341	pImage->LCHSGeometry = *pLCHSGeometry;
1342
1343	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1344	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1345	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1346
1347	/* Create image file. */
1348	fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1349	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1350	if (RT_FAILURE(rc))
1351	{
1352	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1353	goto out;
1354	}
1355
1356	/* Init image state. */
1357	pImage->cbSize = cbSize;
1358	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1359	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1360	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1361	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1362	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1363	pImage->cbBackingFilename = 0;
1364	pImage->offBackingFilename = 0;
1365	qedTableMasksInit(pImage);
1366
1367	/* Init L1 table. */
1368	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1369	if (!pImage->paL1Table)
1370	{
1371	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1372	pImage->pszFilename);
1373	goto out;
1374	}
1375
1376	rc = qedL2TblCacheCreate(pImage);
1377	if (RT_FAILURE(rc))
1378	{
1379	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1380	pImage->pszFilename);
1381	goto out;
1382	}
1383
1384	if (RT_SUCCESS(rc) && pfnProgress)
1385	pfnProgress(pvUser, uPercentStart + uPercentSpan * 98 / 100);
1386
1387	rc = qedFlushImage(pImage);
1388
1389	out:
1390	if (RT_SUCCESS(rc) && pfnProgress)
1391	pfnProgress(pvUser, uPercentStart + uPercentSpan);
1392
1393	if (RT_FAILURE(rc))
1394	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1395	return rc;
1396	}
1397
1398	/**
1399	* Rollback anything done during async cluster allocation.
1400	*
1401	* @returns VBox status code.
1402	* @param pImage The image instance data.
1403	* @param pIoCtx The I/O context.
1404	* @param pClusterAlloc The cluster allocation to rollback.
1405	*/
1406	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1407	{
1408	RT_NOREF1(pIoCtx);
1409	int rc = VINF_SUCCESS;
1410
1411	switch (pClusterAlloc->enmAllocState)
1412	{
1413	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1414	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1415	{
1416	/* Assumption right now is that the L1 table is not modified if the link fails. */
1417	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1418	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1419	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1420	break;
1421	}
1422	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1423	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1424	{
1425	/* Assumption right now is that the L2 table is not modified if the link fails. */
1426	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1427	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1428	break;
1429	}
1430	default:
1431	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1432	rc = VERR_INVALID_STATE;
1433	}
1434
1435	RTMemFree(pClusterAlloc);
1436	return rc;
1437	}
1438
1439	/**
1440	* Updates the state of the async cluster allocation.
1441	*
1442	* @returns VBox status code.
1443	* @param pBackendData The opaque backend data.
1444	* @param pIoCtx I/O context associated with this request.
1445	* @param pvUser Opaque user data passed during a read/write request.
1446	* @param rcReq Status code for the completed request.
1447	*/
1448	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1449	{
1450	int rc = VINF_SUCCESS;
1451	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1452	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1453
1454	if (RT_FAILURE(rcReq))
1455	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1456
1457	AssertPtr(pClusterAlloc->pL2Entry);
1458
1459	switch (pClusterAlloc->enmAllocState)
1460	{
1461	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1462	{
1463	uint64_t offUpdateLe = RT_H2LE_U64(pClusterAlloc->pL2Entry->offL2Tbl);
1464
1465	/* Update the link in the on disk L1 table now. */
1466	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1467	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1468	pImage->offL1Table + pClusterAlloc->idxL1*sizeof(uint64_t),
1469	&offUpdateLe, sizeof(uint64_t), pIoCtx,
1470	qedAsyncClusterAllocUpdate, pClusterAlloc);
1471	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1472	break;
1473	else if (RT_FAILURE(rc))
1474	{
1475	/* Rollback. */
1476	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1477	break;
1478	}
1479	/* Success, fall through. */
1480	}
1481	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1482	{
1483	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1484	uint64_t offData = qedClusterAllocate(pImage, 1);
1485
1486	/* Update the link in the in memory L1 table now. */
1487	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1488	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1489
1490	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1491	pClusterAlloc->cbImageOld = offData;
1492	pClusterAlloc->offClusterNew = offData;
1493
1494	/* Write data. */
1495	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1496	offData, pIoCtx, pClusterAlloc->cbToWrite,
1497	qedAsyncClusterAllocUpdate, pClusterAlloc);
1498	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1499	break;
1500	else if (RT_FAILURE(rc))
1501	{
1502	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1503	RTMemFree(pClusterAlloc);
1504	break;
1505	}
1506	}
1507	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1508	{
1509	uint64_t offUpdateLe = RT_H2LE_U64(pClusterAlloc->offClusterNew);
1510
1511	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1512
1513	/* Link L2 table and update it. */
1514	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1515	pImage->paL1Table[pClusterAlloc->idxL1] + pClusterAlloc->idxL2*sizeof(uint64_t),
1516	&offUpdateLe, sizeof(uint64_t), pIoCtx,
1517	qedAsyncClusterAllocUpdate, pClusterAlloc);
1518	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1519	break;
1520	else if (RT_FAILURE(rc))
1521	{
1522	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1523	RTMemFree(pClusterAlloc);
1524	break;
1525	}
1526	}
1527	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1528	{
1529	/* Everything done without errors, signal completion. */
1530	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1531	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1532	RTMemFree(pClusterAlloc);
1533	rc = VINF_SUCCESS;
1534	break;
1535	}
1536	default:
1537	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1538	pClusterAlloc->enmAllocState));
1539	}
1540
1541	return rc;
1542	}
1543
1544	/** @copydoc VBOXHDDBACKEND::pfnCheckIfValid */
1545	static DECLCALLBACK(int) qedCheckIfValid(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1546	PVDINTERFACE pVDIfsImage, VDTYPE *penmType)
1547	{
1548	RT_NOREF1(pVDIfsDisk);
1549	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1550	PVDIOSTORAGE pStorage = NULL;
1551	uint64_t cbFile;
1552	int rc = VINF_SUCCESS;
1553
1554	/* Get I/O interface. */
1555	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1556	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1557
1558	if ( !VALID_PTR(pszFilename)
1559	\|\| !*pszFilename)
1560	{
1561	rc = VERR_INVALID_PARAMETER;
1562	goto out;
1563	}
1564
1565	/*
1566	* Open the file and read the footer.
1567	*/
1568	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1569	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1570	false /* fCreate */),
1571	&pStorage);
1572	if (RT_SUCCESS(rc))
1573	{
1574	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1575	if ( RT_SUCCESS(rc)
1576	&& cbFile > sizeof(QedHeader))
1577	{
1578	QedHeader Header;
1579
1580	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1581	if ( RT_SUCCESS(rc)
1582	&& qedHdrConvertToHostEndianess(&Header))
1583	{
1584	*penmType = VDTYPE_HDD;
1585	rc = VINF_SUCCESS;
1586	}
1587	else
1588	rc = VERR_VD_GEN_INVALID_HEADER;
1589	}
1590	else
1591	rc = VERR_VD_GEN_INVALID_HEADER;
1592	}
1593
1594	if (pStorage)
1595	vdIfIoIntFileClose(pIfIo, pStorage);
1596
1597	out:
1598	LogFlowFunc(("returns %Rrc\n", rc));
1599	return rc;
1600	}
1601
1602	/** @copydoc VBOXHDDBACKEND::pfnOpen */
1603	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1604	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1605	VDTYPE enmType, void **ppBackendData)
1606	{
1607	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n", pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1608	int rc;
1609	PQEDIMAGE pImage;
1610
1611	NOREF(enmType); /*< @todo r=klaus make use of the type info. /
1612
1613	/* Check open flags. All valid flags are supported. */
1614	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1615	{
1616	rc = VERR_INVALID_PARAMETER;
1617	goto out;
1618	}
1619
1620	/* Check remaining arguments. */
1621	if ( !VALID_PTR(pszFilename)
1622	\|\| !*pszFilename)
1623	{
1624	rc = VERR_INVALID_PARAMETER;
1625	goto out;
1626	}
1627
1628
1629	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1630	if (!pImage)
1631	{
1632	rc = VERR_NO_MEMORY;
1633	goto out;
1634	}
1635	pImage->pszFilename = pszFilename;
1636	pImage->pStorage = NULL;
1637	pImage->pVDIfsDisk = pVDIfsDisk;
1638	pImage->pVDIfsImage = pVDIfsImage;
1639
1640	rc = qedOpenImage(pImage, uOpenFlags);
1641	if (RT_SUCCESS(rc))
1642	*ppBackendData = pImage;
1643	else
1644	RTMemFree(pImage);
1645
1646	out:
1647	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1648	return rc;
1649	}
1650
1651	/** @copydoc VBOXHDDBACKEND::pfnCreate */
1652	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1653	unsigned uImageFlags, const char *pszComment,
1654	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1655	PCRTUUID pUuid, unsigned uOpenFlags,
1656	unsigned uPercentStart, unsigned uPercentSpan,
1657	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1658	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1659	void **ppBackendData)
1660	{
1661	RT_NOREF1(pUuid);
1662	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",
1663	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1664	int rc;
1665	PQEDIMAGE pImage;
1666
1667	PFNVDPROGRESS pfnProgress = NULL;
1668	void *pvUser = NULL;
1669	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1670	if (pIfProgress)
1671	{
1672	pfnProgress = pIfProgress->pfnProgress;
1673	pvUser = pIfProgress->Core.pvUser;
1674	}
1675
1676	/* Check the VD container type. */
1677	if (enmType != VDTYPE_HDD)
1678	{
1679	rc = VERR_VD_INVALID_TYPE;
1680	goto out;
1681	}
1682
1683	/* Check open flags. All valid flags are supported. */
1684	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1685	{
1686	rc = VERR_INVALID_PARAMETER;
1687	goto out;
1688	}
1689
1690	/* Check remaining arguments. */
1691	if ( !VALID_PTR(pszFilename)
1692	\|\| !*pszFilename
1693	\|\| !VALID_PTR(pPCHSGeometry)
1694	\|\| !VALID_PTR(pLCHSGeometry))
1695	{
1696	rc = VERR_INVALID_PARAMETER;
1697	goto out;
1698	}
1699
1700	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1701	if (!pImage)
1702	{
1703	rc = VERR_NO_MEMORY;
1704	goto out;
1705	}
1706	pImage->pszFilename = pszFilename;
1707	pImage->pStorage = NULL;
1708	pImage->pVDIfsDisk = pVDIfsDisk;
1709	pImage->pVDIfsImage = pVDIfsImage;
1710
1711	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1712	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1713	pfnProgress, pvUser, uPercentStart, uPercentSpan);
1714	if (RT_SUCCESS(rc))
1715	{
1716	/* So far the image is opened in read/write mode. Make sure the
1717	* image is opened in read-only mode if the caller requested that. */
1718	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1719	{
1720	qedFreeImage(pImage, false);
1721	rc = qedOpenImage(pImage, uOpenFlags);
1722	if (RT_FAILURE(rc))
1723	{
1724	RTMemFree(pImage);
1725	goto out;
1726	}
1727	}
1728	*ppBackendData = pImage;
1729	}
1730	else
1731	RTMemFree(pImage);
1732
1733	out:
1734	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1735	return rc;
1736	}
1737
1738	/** @copydoc VBOXHDDBACKEND::pfnRename */
1739	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1740	{
1741	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1742	int rc = VINF_SUCCESS;
1743	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1744
1745	/* Check arguments. */
1746	if ( !pImage
1747	\|\| !pszFilename
1748	\|\| !*pszFilename)
1749	{
1750	rc = VERR_INVALID_PARAMETER;
1751	goto out;
1752	}
1753
1754	/* Close the image. */
1755	rc = qedFreeImage(pImage, false);
1756	if (RT_FAILURE(rc))
1757	goto out;
1758
1759	/* Rename the file. */
1760	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1761	if (RT_FAILURE(rc))
1762	{
1763	/* The move failed, try to reopen the original image. */
1764	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1765	if (RT_FAILURE(rc2))
1766	rc = rc2;
1767
1768	goto out;
1769	}
1770
1771	/* Update pImage with the new information. */
1772	pImage->pszFilename = pszFilename;
1773
1774	/* Open the old image with new name. */
1775	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1776	if (RT_FAILURE(rc))
1777	goto out;
1778
1779	out:
1780	LogFlowFunc(("returns %Rrc\n", rc));
1781	return rc;
1782	}
1783
1784	/** @copydoc VBOXHDDBACKEND::pfnClose */
1785	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1786	{
1787	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1788	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1789	int rc;
1790
1791	rc = qedFreeImage(pImage, fDelete);
1792	RTMemFree(pImage);
1793
1794	LogFlowFunc(("returns %Rrc\n", rc));
1795	return rc;
1796	}
1797
1798	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1799	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1800	{
1801	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1802	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1803	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1804	uint32_t offCluster = 0;
1805	uint32_t idxL1 = 0;
1806	uint32_t idxL2 = 0;
1807	uint64_t offFile = 0;
1808	int rc;
1809
1810	AssertPtr(pImage);
1811	Assert(uOffset % 512 == 0);
1812	Assert(cbToRead % 512 == 0);
1813
1814	if (!VALID_PTR(pIoCtx) \|\| !cbToRead)
1815	{
1816	rc = VERR_INVALID_PARAMETER;
1817	goto out;
1818	}
1819
1820	if ( uOffset + cbToRead > pImage->cbSize
1821	\|\| cbToRead == 0)
1822	{
1823	rc = VERR_INVALID_PARAMETER;
1824	goto out;
1825	}
1826
1827	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1828
1829	/* Clip read size to remain in the cluster. */
1830	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1831
1832	/* Get offset in image. */
1833	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1834	if (RT_SUCCESS(rc))
1835	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1836	pIoCtx, cbToRead);
1837
1838	if ( ( RT_SUCCESS(rc)
1839	\|\| rc == VERR_VD_BLOCK_FREE
1840	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1841	&& pcbActuallyRead)
1842	*pcbActuallyRead = cbToRead;
1843
1844	out:
1845	LogFlowFunc(("returns %Rrc\n", rc));
1846	return rc;
1847	}
1848
1849	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1850	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1851	size_t *pcbPostRead, unsigned fWrite)
1852	{
1853	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1854	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1855	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1856	uint32_t offCluster = 0;
1857	uint32_t idxL1 = 0;
1858	uint32_t idxL2 = 0;
1859	uint64_t offImage = 0;
1860	int rc = VINF_SUCCESS;
1861
1862	AssertPtr(pImage);
1863	Assert(!(uOffset % 512));
1864	Assert(!(cbToWrite % 512));
1865
1866	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
1867	{
1868	rc = VERR_VD_IMAGE_READ_ONLY;
1869	goto out;
1870	}
1871
1872	if (!VALID_PTR(pIoCtx) \|\| !cbToWrite)
1873	{
1874	rc = VERR_INVALID_PARAMETER;
1875	goto out;
1876	}
1877
1878	if ( uOffset + cbToWrite > pImage->cbSize
1879	\|\| cbToWrite == 0)
1880	{
1881	rc = VERR_INVALID_PARAMETER;
1882	goto out;
1883	}
1884
1885	/* Convert offset to L1, L2 index and cluster offset. */
1886	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1887
1888	/* Clip write size to remain in the cluster. */
1889	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1890	Assert(!(cbToWrite % 512));
1891
1892	/* Get offset in image. */
1893	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1894	if (RT_SUCCESS(rc))
1895	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1896	offImage, pIoCtx, cbToWrite, NULL, NULL);
1897	else if (rc == VERR_VD_BLOCK_FREE)
1898	{
1899	if ( cbToWrite == pImage->cbCluster
1900	&& !(fWrite & VD_WRITE_NO_ALLOC))
1901	{
1902	PQEDL2CACHEENTRY pL2Entry = NULL;
1903
1904	/* Full cluster write to previously unallocated cluster.
1905	* Allocate cluster and write data. */
1906	Assert(!offCluster);
1907
1908	do
1909	{
1910	/* Check if we have to allocate a new cluster for L2 tables. */
1911	if (!pImage->paL1Table[idxL1])
1912	{
1913	uint64_t offL2Tbl;
1914	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1915
1916	/* Allocate new async cluster allocation state. */
1917	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1918	if (RT_UNLIKELY(!pL2ClusterAlloc))
1919	{
1920	rc = VERR_NO_MEMORY;
1921	break;
1922	}
1923
1924	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1925	if (!pL2Entry)
1926	{
1927	rc = VERR_NO_MEMORY;
1928	RTMemFree(pL2ClusterAlloc);
1929	break;
1930	}
1931
1932	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1933	pL2Entry->offL2Tbl = offL2Tbl;
1934	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1935
1936	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1937	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1938	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1939	pL2ClusterAlloc->idxL1 = idxL1;
1940	pL2ClusterAlloc->idxL2 = idxL2;
1941	pL2ClusterAlloc->cbToWrite = cbToWrite;
1942	pL2ClusterAlloc->pL2Entry = pL2Entry;
1943
1944	/*
1945	* Write the L2 table first and link to the L1 table afterwards.
1946	* If something unexpected happens the worst case which can happen
1947	* is a leak of some clusters.
1948	*/
1949	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1950	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1951	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1952	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1953	break;
1954	else if (RT_FAILURE(rc))
1955	{
1956	RTMemFree(pL2ClusterAlloc);
1957	qedL2TblCacheEntryFree(pImage, pL2Entry);
1958	break;
1959	}
1960
1961	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1962	}
1963	else
1964	{
1965	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1966	&pL2Entry);
1967
1968	if (RT_SUCCESS(rc))
1969	{
1970	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1971
1972	/* Allocate new async cluster allocation state. */
1973	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1974	if (RT_UNLIKELY(!pDataClusterAlloc))
1975	{
1976	rc = VERR_NO_MEMORY;
1977	break;
1978	}
1979
1980	/* Allocate new cluster for the data. */
1981	uint64_t offData = qedClusterAllocate(pImage, 1);
1982
1983	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1984	pDataClusterAlloc->cbImageOld = offData;
1985	pDataClusterAlloc->offClusterNew = offData;
1986	pDataClusterAlloc->idxL1 = idxL1;
1987	pDataClusterAlloc->idxL2 = idxL2;
1988	pDataClusterAlloc->cbToWrite = cbToWrite;
1989	pDataClusterAlloc->pL2Entry = pL2Entry;
1990
1991	/* Write data. */
1992	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1993	offData, pIoCtx, cbToWrite,
1994	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1995	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1996	break;
1997	else if (RT_FAILURE(rc))
1998	{
1999	RTMemFree(pDataClusterAlloc);
2000	break;
2001	}
2002
2003	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
2004	}
2005	}
2006
2007	} while (0);
2008
2009	*pcbPreRead = 0;
2010	*pcbPostRead = 0;
2011	}
2012	else
2013	{
2014	/* Trying to do a partial write to an unallocated cluster. Don't do
2015	* anything except letting the upper layer know what to do. */
2016	*pcbPreRead = offCluster;
2017	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
2018	}
2019	}
2020
2021	if (pcbWriteProcess)
2022	*pcbWriteProcess = cbToWrite;
2023
2024
2025	out:
2026	LogFlowFunc(("returns %Rrc\n", rc));
2027	return rc;
2028	}
2029
2030	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
2031	{
2032	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2033	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2034	int rc = VINF_SUCCESS;
2035
2036	Assert(pImage);
2037
2038	if (VALID_PTR(pIoCtx))
2039	rc = qedFlushImageAsync(pImage, pIoCtx);
2040	else
2041	rc = VERR_INVALID_PARAMETER;
2042
2043	LogFlowFunc(("returns %Rrc\n", rc));
2044	return rc;
2045	}
2046
2047	/** @copydoc VBOXHDDBACKEND::pfnGetVersion */
2048	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
2049	{
2050	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2051	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2052
2053	AssertPtr(pImage);
2054
2055	if (pImage)
2056	return 1;
2057	else
2058	return 0;
2059	}
2060
2061	/** @copydoc VBOXHDDBACKEND::pfnGetSectorSize */
2062	static DECLCALLBACK(uint32_t) qedGetSectorSize(void *pBackendData)
2063	{
2064	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2065	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2066	uint32_t cb = 0;
2067
2068	AssertPtr(pImage);
2069
2070	if (pImage && pImage->pStorage)
2071	cb = 512;
2072
2073	LogFlowFunc(("returns %u\n", cb));
2074	return cb;
2075	}
2076
2077	/** @copydoc VBOXHDDBACKEND::pfnGetSize */
2078	static DECLCALLBACK(uint64_t) qedGetSize(void *pBackendData)
2079	{
2080	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2081	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2082	uint64_t cb = 0;
2083
2084	AssertPtr(pImage);
2085
2086	if (pImage && pImage->pStorage)
2087	cb = pImage->cbSize;
2088
2089	LogFlowFunc(("returns %llu\n", cb));
2090	return cb;
2091	}
2092
2093	/** @copydoc VBOXHDDBACKEND::pfnGetFileSize */
2094	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
2095	{
2096	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2097	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2098	uint64_t cb = 0;
2099
2100	AssertPtr(pImage);
2101
2102	if (pImage)
2103	{
2104	uint64_t cbFile;
2105	if (pImage->pStorage)
2106	{
2107	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
2108	if (RT_SUCCESS(rc))
2109	cb += cbFile;
2110	}
2111	}
2112
2113	LogFlowFunc(("returns %lld\n", cb));
2114	return cb;
2115	}
2116
2117	/** @copydoc VBOXHDDBACKEND::pfnGetPCHSGeometry */
2118	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
2119	PVDGEOMETRY pPCHSGeometry)
2120	{
2121	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
2122	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2123	int rc;
2124
2125	AssertPtr(pImage);
2126
2127	if (pImage)
2128	{
2129	if (pImage->PCHSGeometry.cCylinders)
2130	{
2131	*pPCHSGeometry = pImage->PCHSGeometry;
2132	rc = VINF_SUCCESS;
2133	}
2134	else
2135	rc = VERR_VD_GEOMETRY_NOT_SET;
2136	}
2137	else
2138	rc = VERR_VD_NOT_OPENED;
2139
2140	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2141	return rc;
2142	}
2143
2144	/** @copydoc VBOXHDDBACKEND::pfnSetPCHSGeometry */
2145	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
2146	PCVDGEOMETRY pPCHSGeometry)
2147	{
2148	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n", pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2149	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2150	int rc;
2151
2152	AssertPtr(pImage);
2153
2154	if (pImage)
2155	{
2156	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2157	{
2158	rc = VERR_VD_IMAGE_READ_ONLY;
2159	goto out;
2160	}
2161
2162	pImage->PCHSGeometry = *pPCHSGeometry;
2163	rc = VINF_SUCCESS;
2164	}
2165	else
2166	rc = VERR_VD_NOT_OPENED;
2167
2168	out:
2169	LogFlowFunc(("returns %Rrc\n", rc));
2170	return rc;
2171	}
2172
2173	/** @copydoc VBOXHDDBACKEND::pfnGetLCHSGeometry */
2174	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData,
2175	PVDGEOMETRY pLCHSGeometry)
2176	{
2177	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2178	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2179	int rc;
2180
2181	AssertPtr(pImage);
2182
2183	if (pImage)
2184	{
2185	if (pImage->LCHSGeometry.cCylinders)
2186	{
2187	*pLCHSGeometry = pImage->LCHSGeometry;
2188	rc = VINF_SUCCESS;
2189	}
2190	else
2191	rc = VERR_VD_GEOMETRY_NOT_SET;
2192	}
2193	else
2194	rc = VERR_VD_NOT_OPENED;
2195
2196	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2197	return rc;
2198	}
2199
2200	/** @copydoc VBOXHDDBACKEND::pfnSetLCHSGeometry */
2201	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData,
2202	PCVDGEOMETRY pLCHSGeometry)
2203	{
2204	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData, pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2205	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2206	int rc;
2207
2208	AssertPtr(pImage);
2209
2210	if (pImage)
2211	{
2212	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2213	{
2214	rc = VERR_VD_IMAGE_READ_ONLY;
2215	goto out;
2216	}
2217
2218	pImage->LCHSGeometry = *pLCHSGeometry;
2219	rc = VINF_SUCCESS;
2220	}
2221	else
2222	rc = VERR_VD_NOT_OPENED;
2223
2224	out:
2225	LogFlowFunc(("returns %Rrc\n", rc));
2226	return rc;
2227	}
2228
2229	/** @copydoc VBOXHDDBACKEND::pfnGetImageFlags */
2230	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2231	{
2232	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2233	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2234	unsigned uImageFlags;
2235
2236	AssertPtr(pImage);
2237
2238	if (pImage)
2239	uImageFlags = pImage->uImageFlags;
2240	else
2241	uImageFlags = 0;
2242
2243	LogFlowFunc(("returns %#x\n", uImageFlags));
2244	return uImageFlags;
2245	}
2246
2247	/** @copydoc VBOXHDDBACKEND::pfnGetOpenFlags */
2248	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2249	{
2250	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2251	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2252	unsigned uOpenFlags;
2253
2254	AssertPtr(pImage);
2255
2256	if (pImage)
2257	uOpenFlags = pImage->uOpenFlags;
2258	else
2259	uOpenFlags = 0;
2260
2261	LogFlowFunc(("returns %#x\n", uOpenFlags));
2262	return uOpenFlags;
2263	}
2264
2265	/** @copydoc VBOXHDDBACKEND::pfnSetOpenFlags */
2266	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2267	{
2268	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2269	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2270	int rc;
2271
2272	/* Image must be opened and the new flags must be valid. */
2273	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2274	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2275	{
2276	rc = VERR_INVALID_PARAMETER;
2277	goto out;
2278	}
2279
2280	/* Implement this operation via reopening the image. */
2281	rc = qedFreeImage(pImage, false);
2282	if (RT_FAILURE(rc))
2283	goto out;
2284	rc = qedOpenImage(pImage, uOpenFlags);
2285
2286	out:
2287	LogFlowFunc(("returns %Rrc\n", rc));
2288	return rc;
2289	}
2290
2291	/** @copydoc VBOXHDDBACKEND::pfnGetComment */
2292	static DECLCALLBACK(int) qedGetComment(void pBackendData, char pszComment,
2293	size_t cbComment)
2294	{
2295	RT_NOREF2(pszComment, cbComment);
2296	LogFlowFunc(("pBackendData=%#p pszComment=%#p cbComment=%zu\n", pBackendData, pszComment, cbComment));
2297	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2298	int rc;
2299
2300	AssertPtr(pImage);
2301
2302	if (pImage)
2303	rc = VERR_NOT_SUPPORTED;
2304	else
2305	rc = VERR_VD_NOT_OPENED;
2306
2307	LogFlowFunc(("returns %Rrc comment='%s'\n", rc, pszComment));
2308	return rc;
2309	}
2310
2311	/** @copydoc VBOXHDDBACKEND::pfnSetComment */
2312	static DECLCALLBACK(int) qedSetComment(void pBackendData, const char pszComment)
2313	{
2314	RT_NOREF1(pszComment);
2315	LogFlowFunc(("pBackendData=%#p pszComment=\"%s\"\n", pBackendData, pszComment));
2316	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2317	int rc;
2318
2319	AssertPtr(pImage);
2320
2321	if (pImage)
2322	{
2323	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2324	rc = VERR_VD_IMAGE_READ_ONLY;
2325	else
2326	rc = VERR_NOT_SUPPORTED;
2327	}
2328	else
2329	rc = VERR_VD_NOT_OPENED;
2330
2331	LogFlowFunc(("returns %Rrc\n", rc));
2332	return rc;
2333	}
2334
2335	/** @copydoc VBOXHDDBACKEND::pfnGetUuid */
2336	static DECLCALLBACK(int) qedGetUuid(void *pBackendData, PRTUUID pUuid)
2337	{
2338	RT_NOREF1(pUuid);
2339	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2340	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2341	int rc;
2342
2343	AssertPtr(pImage);
2344
2345	if (pImage)
2346	rc = VERR_NOT_SUPPORTED;
2347	else
2348	rc = VERR_VD_NOT_OPENED;
2349
2350	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2351	return rc;
2352	}
2353
2354	/** @copydoc VBOXHDDBACKEND::pfnSetUuid */
2355	static DECLCALLBACK(int) qedSetUuid(void *pBackendData, PCRTUUID pUuid)
2356	{
2357	RT_NOREF1(pUuid);
2358	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2359	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2360	int rc;
2361
2362	LogFlowFunc(("%RTuuid\n", pUuid));
2363	AssertPtr(pImage);
2364
2365	if (pImage)
2366	{
2367	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2368	rc = VERR_NOT_SUPPORTED;
2369	else
2370	rc = VERR_VD_IMAGE_READ_ONLY;
2371	}
2372	else
2373	rc = VERR_VD_NOT_OPENED;
2374
2375	LogFlowFunc(("returns %Rrc\n", rc));
2376	return rc;
2377	}
2378
2379	/** @copydoc VBOXHDDBACKEND::pfnGetModificationUuid */
2380	static DECLCALLBACK(int) qedGetModificationUuid(void *pBackendData, PRTUUID pUuid)
2381	{
2382	RT_NOREF1(pUuid);
2383	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2384	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2385	int rc;
2386
2387	AssertPtr(pImage);
2388
2389	if (pImage)
2390	rc = VERR_NOT_SUPPORTED;
2391	else
2392	rc = VERR_VD_NOT_OPENED;
2393
2394	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2395	return rc;
2396	}
2397
2398	/** @copydoc VBOXHDDBACKEND::pfnSetModificationUuid */
2399	static DECLCALLBACK(int) qedSetModificationUuid(void *pBackendData, PCRTUUID pUuid)
2400	{
2401	RT_NOREF1(pUuid);
2402	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2403	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2404	int rc;
2405
2406	AssertPtr(pImage);
2407
2408	if (pImage)
2409	{
2410	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2411	rc = VERR_NOT_SUPPORTED;
2412	else
2413	rc = VERR_VD_IMAGE_READ_ONLY;
2414	}
2415	else
2416	rc = VERR_VD_NOT_OPENED;
2417
2418	LogFlowFunc(("returns %Rrc\n", rc));
2419	return rc;
2420	}
2421
2422	/** @copydoc VBOXHDDBACKEND::pfnGetParentUuid */
2423	static DECLCALLBACK(int) qedGetParentUuid(void *pBackendData, PRTUUID pUuid)
2424	{
2425	RT_NOREF1(pUuid);
2426	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2427	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2428	int rc;
2429
2430	AssertPtr(pImage);
2431
2432	if (pImage)
2433	rc = VERR_NOT_SUPPORTED;
2434	else
2435	rc = VERR_VD_NOT_OPENED;
2436
2437	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2438	return rc;
2439	}
2440
2441	/** @copydoc VBOXHDDBACKEND::pfnSetParentUuid */
2442	static DECLCALLBACK(int) qedSetParentUuid(void *pBackendData, PCRTUUID pUuid)
2443	{
2444	RT_NOREF1(pUuid);
2445	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2446	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2447	int rc;
2448
2449	AssertPtr(pImage);
2450
2451	if (pImage)
2452	{
2453	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2454	rc = VERR_NOT_SUPPORTED;
2455	else
2456	rc = VERR_VD_IMAGE_READ_ONLY;
2457	}
2458	else
2459	rc = VERR_VD_NOT_OPENED;
2460
2461	LogFlowFunc(("returns %Rrc\n", rc));
2462	return rc;
2463	}
2464
2465	/** @copydoc VBOXHDDBACKEND::pfnGetParentModificationUuid */
2466	static DECLCALLBACK(int) qedGetParentModificationUuid(void *pBackendData, PRTUUID pUuid)
2467	{
2468	RT_NOREF1(pUuid);
2469	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2470	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2471	int rc;
2472
2473	AssertPtr(pImage);
2474
2475	if (pImage)
2476	rc = VERR_NOT_SUPPORTED;
2477	else
2478	rc = VERR_VD_NOT_OPENED;
2479
2480	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2481	return rc;
2482	}
2483
2484	/** @copydoc VBOXHDDBACKEND::pfnSetParentModificationUuid */
2485	static DECLCALLBACK(int) qedSetParentModificationUuid(void *pBackendData, PCRTUUID pUuid)
2486	{
2487	RT_NOREF1(pUuid);
2488	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2489	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2490	int rc;
2491
2492	AssertPtr(pImage);
2493
2494	if (pImage)
2495	{
2496	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2497	rc = VERR_NOT_SUPPORTED;
2498	else
2499	rc = VERR_VD_IMAGE_READ_ONLY;
2500	}
2501	else
2502	rc = VERR_VD_NOT_OPENED;
2503
2504	LogFlowFunc(("returns %Rrc\n", rc));
2505	return rc;
2506	}
2507
2508	/** @copydoc VBOXHDDBACKEND::pfnDump */
2509	static DECLCALLBACK(void) qedDump(void *pBackendData)
2510	{
2511	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2512
2513	AssertPtr(pImage);
2514	if (pImage)
2515	{
2516	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2517	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2518	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2519	pImage->cbSize / 512);
2520	}
2521	}
2522
2523	/** @copydoc VBOXHDDBACKEND::pfnGetParentFilename */
2524	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2525	{
2526	int rc = VINF_SUCCESS;
2527	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2528
2529	AssertPtr(pImage);
2530	if (pImage)
2531	if (pImage->pszBackingFilename)
2532	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2533	else
2534	rc = VERR_NOT_SUPPORTED;
2535	else
2536	rc = VERR_VD_NOT_OPENED;
2537
2538	LogFlowFunc(("returns %Rrc\n", rc));
2539	return rc;
2540	}
2541
2542	/** @copydoc VBOXHDDBACKEND::pfnSetParentFilename */
2543	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2544	{
2545	int rc = VINF_SUCCESS;
2546	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2547
2548	AssertPtr(pImage);
2549	if (pImage)
2550	{
2551	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2552	rc = VERR_VD_IMAGE_READ_ONLY;
2553	else if ( pImage->pszBackingFilename
2554	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2555	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2556	else
2557	{
2558	if (pImage->pszBackingFilename)
2559	RTStrFree(pImage->pszBackingFilename);
2560	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2561	if (!pImage->pszBackingFilename)
2562	rc = VERR_NO_MEMORY;
2563	else
2564	{
2565	if (!pImage->offBackingFilename)
2566	{
2567	/* Allocate new cluster. */
2568	uint64_t offData = qedClusterAllocate(pImage, 1);
2569
2570	Assert((offData & UINT32_MAX) == offData);
2571	pImage->offBackingFilename = (uint32_t)offData;
2572	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2573	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2574	offData + pImage->cbCluster);
2575	}
2576
2577	if (RT_SUCCESS(rc))
2578	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2579	pImage->offBackingFilename,
2580	pImage->pszBackingFilename,
2581	strlen(pImage->pszBackingFilename));
2582	}
2583	}
2584	}
2585	else
2586	rc = VERR_VD_NOT_OPENED;
2587
2588	LogFlowFunc(("returns %Rrc\n", rc));
2589	return rc;
2590	}
2591
2592	/** @copydoc VBOXHDDBACKEND::pfnResize */
2593	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2594	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2595	unsigned uPercentStart, unsigned uPercentSpan,
2596	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2597	PVDINTERFACE pVDIfsOperation)
2598	{
2599	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2600	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2601	int rc = VINF_SUCCESS;
2602
2603	/* Making the image smaller is not supported at the moment. */
2604	if (cbSize < pImage->cbSize)
2605	rc = VERR_NOT_SUPPORTED;
2606	else if (cbSize > pImage->cbSize)
2607	{
2608	/*
2609	* It is enough to just update the size field in the header to complete
2610	* growing. With the default cluster and table sizes the image can be expanded
2611	* to 64TB without overflowing the L1 and L2 tables making block relocation
2612	* superfluous.
2613	* @todo: The rare case where block relocation is still required (non default
2614	* table and/or cluster size or images with more than 64TB) is not
2615	* implemented yet and resizing such an image will fail with an error.
2616	*/
2617	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2618	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2619	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2620	pImage->pszFilename);
2621	else
2622	{
2623	uint64_t cbSizeOld = pImage->cbSize;
2624
2625	pImage->cbSize = cbSize;
2626	rc = qedFlushImage(pImage);
2627	if (RT_FAILURE(rc))
2628	{
2629	pImage->cbSize = cbSizeOld; /* Restore */
2630
2631	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2632	pImage->pszFilename);
2633	}
2634	}
2635	}
2636	/* Same size doesn't change the image at all. */
2637
2638	LogFlowFunc(("returns %Rrc\n", rc));
2639	return rc;
2640	}
2641
2642
2643	const VBOXHDDBACKEND g_QedBackend =
2644	{
2645	/* pszBackendName */
2646	"QED",
2647	/* cbSize */
2648	sizeof(VBOXHDDBACKEND),
2649	/* uBackendCaps */
2650	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2651	/* paFileExtensions */
2652	s_aQedFileExtensions,
2653	/* paConfigInfo */
2654	NULL,
2655	/* pfnCheckIfValid */
2656	qedCheckIfValid,
2657	/* pfnOpen */
2658	qedOpen,
2659	/* pfnCreate */
2660	qedCreate,
2661	/* pfnRename */
2662	qedRename,
2663	/* pfnClose */
2664	qedClose,
2665	/* pfnRead */
2666	qedRead,
2667	/* pfnWrite */
2668	qedWrite,
2669	/* pfnFlush */
2670	qedFlush,
2671	/* pfnDiscard */
2672	NULL,
2673	/* pfnGetVersion */
2674	qedGetVersion,
2675	/* pfnGetSectorSize */
2676	qedGetSectorSize,
2677	/* pfnGetSize */
2678	qedGetSize,
2679	/* pfnGetFileSize */
2680	qedGetFileSize,
2681	/* pfnGetPCHSGeometry */
2682	qedGetPCHSGeometry,
2683	/* pfnSetPCHSGeometry */
2684	qedSetPCHSGeometry,
2685	/* pfnGetLCHSGeometry */
2686	qedGetLCHSGeometry,
2687	/* pfnSetLCHSGeometry */
2688	qedSetLCHSGeometry,
2689	/* pfnGetImageFlags */
2690	qedGetImageFlags,
2691	/* pfnGetOpenFlags */
2692	qedGetOpenFlags,
2693	/* pfnSetOpenFlags */
2694	qedSetOpenFlags,
2695	/* pfnGetComment */
2696	qedGetComment,
2697	/* pfnSetComment */
2698	qedSetComment,
2699	/* pfnGetUuid */
2700	qedGetUuid,
2701	/* pfnSetUuid */
2702	qedSetUuid,
2703	/* pfnGetModificationUuid */
2704	qedGetModificationUuid,
2705	/* pfnSetModificationUuid */
2706	qedSetModificationUuid,
2707	/* pfnGetParentUuid */
2708	qedGetParentUuid,
2709	/* pfnSetParentUuid */
2710	qedSetParentUuid,
2711	/* pfnGetParentModificationUuid */
2712	qedGetParentModificationUuid,
2713	/* pfnSetParentModificationUuid */
2714	qedSetParentModificationUuid,
2715	/* pfnDump */
2716	qedDump,
2717	/* pfnGetTimestamp */
2718	NULL,
2719	/* pfnGetParentTimestamp */
2720	NULL,
2721	/* pfnSetParentTimestamp */
2722	NULL,
2723	/* pfnGetParentFilename */
2724	qedGetParentFilename,
2725	/* pfnSetParentFilename */
2726	qedSetParentFilename,
2727	/* pfnComposeLocation */
2728	genericFileComposeLocation,
2729	/* pfnComposeName */
2730	genericFileComposeName,
2731	/* pfnCompact */
2732	NULL,
2733	/* pfnResize */
2734	qedResize,
2735	/* pfnRepair */
2736	NULL,
2737	/* pfnTraverseMetadata */
2738	NULL
2739	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 63699

Download in other formats: