QED.cpp@ 63788

Last change on this file since 63788 was 63788, checked in by vboxsync, 8 years ago
Storage: Remove unused/obsolete VDFormats directory (can be resurrected later if required again)
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 63788 2016-09-09 22:27:48Z 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	pImage->uOpenFlags = uOpenFlags;
1154	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1155	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1156	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1157
1158	/*
1159	* Create the L2 cache before opening the image so we can call qedFreeImage()
1160	* even if opening the image file fails.
1161	*/
1162	int rc = qedL2TblCacheCreate(pImage);
1163	if (RT_FAILURE(rc))
1164	{
1165	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1166	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1167	pImage->pszFilename);
1168
1169	goto out;
1170	}
1171
1172	/*
1173	* Open the image.
1174	*/
1175	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename,
1176	VDOpenFlagsToFileOpenFlags(uOpenFlags,
1177	false /* fCreate */),
1178	&pImage->pStorage);
1179	if (RT_FAILURE(rc))
1180	{
1181	/* Do NOT signal an appropriate error here, as the VD layer has the
1182	* choice of retrying the open if it failed. */
1183	goto out;
1184	}
1185
1186	uint64_t cbFile;
1187	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1188	if (RT_FAILURE(rc))
1189	goto out;
1190	if (cbFile > sizeof(QedHeader))
1191	{
1192	QedHeader Header;
1193
1194	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, 0, &Header, sizeof(Header));
1195	if ( RT_SUCCESS(rc)
1196	&& qedHdrConvertToHostEndianess(&Header))
1197	{
1198	if ( !(Header.u64FeatureFlags & ~QED_FEATURE_MASK)
1199	&& !(Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE_NO_PROBE))
1200	{
1201	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1202	{
1203	/* Image needs checking. */
1204	if (!(uOpenFlags & VD_OPEN_FLAGS_READONLY))
1205	rc = qedCheckImage(pImage, &Header);
1206	else
1207	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1208	N_("Qed: Image '%s' needs checking but is opened readonly"),
1209	pImage->pszFilename);
1210	}
1211
1212	if ( RT_SUCCESS(rc)
1213	&& (Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE))
1214	{
1215	/* Load backing filename from image. */
1216	pImage->pszBackingFilename = (char )RTMemAllocZ(Header.u32BackingFilenameSize + 1); / +1 for \0 terminator. */
1217	if (pImage->pszBackingFilename)
1218	{
1219	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1220	pImage->offBackingFilename = Header.u32OffBackingFilename;
1221	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1222	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1223	Header.u32BackingFilenameSize);
1224	}
1225	else
1226	rc = VERR_NO_MEMORY;
1227	}
1228
1229	if (RT_SUCCESS(rc))
1230	{
1231	pImage->cbImage = cbFile;
1232	pImage->cbCluster = Header.u32ClusterSize;
1233	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1234	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1235	pImage->offL1Table = Header.u64OffL1Table;
1236	pImage->cbSize = Header.u64Size;
1237	qedTableMasksInit(pImage);
1238
1239	/* Allocate L1 table. */
1240	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1241	if (pImage->paL1Table)
1242	{
1243	/* Read from the image. */
1244	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1245	pImage->offL1Table, pImage->paL1Table,
1246	pImage->cbTable);
1247	if (RT_SUCCESS(rc))
1248	{
1249	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1250
1251	/* If the consistency check succeeded, clear the flag by flushing the image. */
1252	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1253	rc = qedFlushImage(pImage);
1254	}
1255	else
1256	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1257	N_("Qed: Reading the L1 table for image '%s' failed"),
1258	pImage->pszFilename);
1259	}
1260	else
1261	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1262	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1263	pImage->pszFilename);
1264	}
1265	}
1266	else
1267	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1268	N_("Qed: The image '%s' makes use of unsupported features"),
1269	pImage->pszFilename);
1270	}
1271	else if (RT_SUCCESS(rc))
1272	rc = VERR_VD_GEN_INVALID_HEADER;
1273	}
1274	else
1275	rc = VERR_VD_GEN_INVALID_HEADER;
1276
1277	out:
1278	if (RT_FAILURE(rc))
1279	qedFreeImage(pImage, false);
1280	return rc;
1281	}
1282
1283	/**
1284	* Internal: Create a qed image.
1285	*/
1286	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1287	unsigned uImageFlags, const char *pszComment,
1288	PCVDGEOMETRY pPCHSGeometry,
1289	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1290	PFNVDPROGRESS pfnProgress, void *pvUser,
1291	unsigned uPercentStart, unsigned uPercentSpan)
1292	{
1293	RT_NOREF1(pszComment);
1294	int rc;
1295	int32_t fOpen;
1296
1297	if (uImageFlags & VD_IMAGE_FLAGS_FIXED)
1298	{
1299	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1300	goto out;
1301	}
1302
1303	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1304	pImage->uImageFlags = uImageFlags;
1305	pImage->PCHSGeometry = *pPCHSGeometry;
1306	pImage->LCHSGeometry = *pLCHSGeometry;
1307
1308	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1309	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1310	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1311
1312	/* Create image file. */
1313	fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1314	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1315	if (RT_FAILURE(rc))
1316	{
1317	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1318	goto out;
1319	}
1320
1321	/* Init image state. */
1322	pImage->cbSize = cbSize;
1323	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1324	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1325	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1326	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1327	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1328	pImage->cbBackingFilename = 0;
1329	pImage->offBackingFilename = 0;
1330	qedTableMasksInit(pImage);
1331
1332	/* Init L1 table. */
1333	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1334	if (!pImage->paL1Table)
1335	{
1336	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1337	pImage->pszFilename);
1338	goto out;
1339	}
1340
1341	rc = qedL2TblCacheCreate(pImage);
1342	if (RT_FAILURE(rc))
1343	{
1344	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1345	pImage->pszFilename);
1346	goto out;
1347	}
1348
1349	if (RT_SUCCESS(rc) && pfnProgress)
1350	pfnProgress(pvUser, uPercentStart + uPercentSpan * 98 / 100);
1351
1352	rc = qedFlushImage(pImage);
1353
1354	out:
1355	if (RT_SUCCESS(rc) && pfnProgress)
1356	pfnProgress(pvUser, uPercentStart + uPercentSpan);
1357
1358	if (RT_FAILURE(rc))
1359	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1360	return rc;
1361	}
1362
1363	/**
1364	* Rollback anything done during async cluster allocation.
1365	*
1366	* @returns VBox status code.
1367	* @param pImage The image instance data.
1368	* @param pIoCtx The I/O context.
1369	* @param pClusterAlloc The cluster allocation to rollback.
1370	*/
1371	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1372	{
1373	RT_NOREF1(pIoCtx);
1374	int rc = VINF_SUCCESS;
1375
1376	switch (pClusterAlloc->enmAllocState)
1377	{
1378	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1379	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1380	{
1381	/* Revert the L1 table entry */
1382	pImage->paL1Table[pClusterAlloc->idxL1] = 0;
1383
1384	/* Assumption right now is that the L1 table is not modified on storage if the link fails. */
1385	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1386	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1387	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1388	break;
1389	}
1390	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1391	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1392	{
1393	/* Assumption right now is that the L2 table is not modified if the link fails. */
1394	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = 0;
1395	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1396	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1397	break;
1398	}
1399	default:
1400	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1401	rc = VERR_INVALID_STATE;
1402	}
1403
1404	RTMemFree(pClusterAlloc);
1405	return rc;
1406	}
1407
1408	/**
1409	* Updates the state of the async cluster allocation.
1410	*
1411	* @returns VBox status code.
1412	* @param pBackendData The opaque backend data.
1413	* @param pIoCtx I/O context associated with this request.
1414	* @param pvUser Opaque user data passed during a read/write request.
1415	* @param rcReq Status code for the completed request.
1416	*/
1417	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1418	{
1419	int rc = VINF_SUCCESS;
1420	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1421	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1422
1423	if (RT_FAILURE(rcReq))
1424	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1425
1426	AssertPtr(pClusterAlloc->pL2Entry);
1427
1428	switch (pClusterAlloc->enmAllocState)
1429	{
1430	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1431	{
1432	/* Update the link in the in memory L1 table now. */
1433	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1434
1435	/* Update the link in the on disk L1 table now. */
1436	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1437	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1438	qedAsyncClusterAllocUpdate, pClusterAlloc);
1439	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1440	break;
1441	else if (RT_FAILURE(rc))
1442	{
1443	/* Rollback. */
1444	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1445	break;
1446	}
1447	/* Success, fall through. */
1448	}
1449	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1450	{
1451	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1452	uint64_t offData = qedClusterAllocate(pImage, 1);
1453
1454	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1455
1456	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1457	pClusterAlloc->cbImageOld = offData;
1458	pClusterAlloc->offClusterNew = offData;
1459
1460	/* Write data. */
1461	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1462	offData, pIoCtx, pClusterAlloc->cbToWrite,
1463	qedAsyncClusterAllocUpdate, pClusterAlloc);
1464	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1465	break;
1466	else if (RT_FAILURE(rc))
1467	{
1468	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1469	RTMemFree(pClusterAlloc);
1470	break;
1471	}
1472	}
1473	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1474	{
1475	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1476	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1477
1478	/* Link L2 table and update it. */
1479	rc = qedTblWrite(pImage, pIoCtx, pImage->paL1Table[pClusterAlloc->idxL1],
1480	pClusterAlloc->pL2Entry->paL2Tbl,
1481	qedAsyncClusterAllocUpdate, pClusterAlloc);
1482	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1483	break;
1484	else if (RT_FAILURE(rc))
1485	{
1486	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1487	RTMemFree(pClusterAlloc);
1488	break;
1489	}
1490	}
1491	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1492	{
1493	/* Everything done without errors, signal completion. */
1494	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1495	RTMemFree(pClusterAlloc);
1496	rc = VINF_SUCCESS;
1497	break;
1498	}
1499	default:
1500	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1501	pClusterAlloc->enmAllocState));
1502	}
1503
1504	return rc;
1505	}
1506
1507	/** @copydoc VDIMAGEBACKEND::pfnCheckIfValid */
1508	static DECLCALLBACK(int) qedCheckIfValid(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1509	PVDINTERFACE pVDIfsImage, VDTYPE *penmType)
1510	{
1511	RT_NOREF1(pVDIfsDisk);
1512	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1513	PVDIOSTORAGE pStorage = NULL;
1514	uint64_t cbFile;
1515	int rc = VINF_SUCCESS;
1516
1517	/* Get I/O interface. */
1518	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1519	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1520
1521	if ( !VALID_PTR(pszFilename)
1522	\|\| !*pszFilename)
1523	{
1524	rc = VERR_INVALID_PARAMETER;
1525	goto out;
1526	}
1527
1528	/*
1529	* Open the file and read the footer.
1530	*/
1531	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1532	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1533	false /* fCreate */),
1534	&pStorage);
1535	if (RT_SUCCESS(rc))
1536	{
1537	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1538	if ( RT_SUCCESS(rc)
1539	&& cbFile > sizeof(QedHeader))
1540	{
1541	QedHeader Header;
1542
1543	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1544	if ( RT_SUCCESS(rc)
1545	&& qedHdrConvertToHostEndianess(&Header))
1546	{
1547	*penmType = VDTYPE_HDD;
1548	rc = VINF_SUCCESS;
1549	}
1550	else
1551	rc = VERR_VD_GEN_INVALID_HEADER;
1552	}
1553	else
1554	rc = VERR_VD_GEN_INVALID_HEADER;
1555	}
1556
1557	if (pStorage)
1558	vdIfIoIntFileClose(pIfIo, pStorage);
1559
1560	out:
1561	LogFlowFunc(("returns %Rrc\n", rc));
1562	return rc;
1563	}
1564
1565	/** @copydoc VDIMAGEBACKEND::pfnOpen */
1566	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1567	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1568	VDTYPE enmType, void **ppBackendData)
1569	{
1570	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n", pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1571	int rc;
1572	PQEDIMAGE pImage;
1573
1574	NOREF(enmType); /*< @todo r=klaus make use of the type info. /
1575
1576	/* Check open flags. All valid flags are supported. */
1577	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1578	{
1579	rc = VERR_INVALID_PARAMETER;
1580	goto out;
1581	}
1582
1583	/* Check remaining arguments. */
1584	if ( !VALID_PTR(pszFilename)
1585	\|\| !*pszFilename)
1586	{
1587	rc = VERR_INVALID_PARAMETER;
1588	goto out;
1589	}
1590
1591
1592	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1593	if (!pImage)
1594	{
1595	rc = VERR_NO_MEMORY;
1596	goto out;
1597	}
1598	pImage->pszFilename = pszFilename;
1599	pImage->pStorage = NULL;
1600	pImage->pVDIfsDisk = pVDIfsDisk;
1601	pImage->pVDIfsImage = pVDIfsImage;
1602
1603	rc = qedOpenImage(pImage, uOpenFlags);
1604	if (RT_SUCCESS(rc))
1605	*ppBackendData = pImage;
1606	else
1607	RTMemFree(pImage);
1608
1609	out:
1610	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1611	return rc;
1612	}
1613
1614	/** @copydoc VDIMAGEBACKEND::pfnCreate */
1615	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1616	unsigned uImageFlags, const char *pszComment,
1617	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1618	PCRTUUID pUuid, unsigned uOpenFlags,
1619	unsigned uPercentStart, unsigned uPercentSpan,
1620	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1621	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1622	void **ppBackendData)
1623	{
1624	RT_NOREF1(pUuid);
1625	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",
1626	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1627	int rc;
1628	PQEDIMAGE pImage;
1629
1630	PFNVDPROGRESS pfnProgress = NULL;
1631	void *pvUser = NULL;
1632	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1633	if (pIfProgress)
1634	{
1635	pfnProgress = pIfProgress->pfnProgress;
1636	pvUser = pIfProgress->Core.pvUser;
1637	}
1638
1639	/* Check the VD container type. */
1640	if (enmType != VDTYPE_HDD)
1641	{
1642	rc = VERR_VD_INVALID_TYPE;
1643	goto out;
1644	}
1645
1646	/* Check open flags. All valid flags are supported. */
1647	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1648	{
1649	rc = VERR_INVALID_PARAMETER;
1650	goto out;
1651	}
1652
1653	/* Check remaining arguments. */
1654	if ( !VALID_PTR(pszFilename)
1655	\|\| !*pszFilename
1656	\|\| !VALID_PTR(pPCHSGeometry)
1657	\|\| !VALID_PTR(pLCHSGeometry))
1658	{
1659	rc = VERR_INVALID_PARAMETER;
1660	goto out;
1661	}
1662
1663	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1664	if (!pImage)
1665	{
1666	rc = VERR_NO_MEMORY;
1667	goto out;
1668	}
1669	pImage->pszFilename = pszFilename;
1670	pImage->pStorage = NULL;
1671	pImage->pVDIfsDisk = pVDIfsDisk;
1672	pImage->pVDIfsImage = pVDIfsImage;
1673
1674	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1675	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1676	pfnProgress, pvUser, uPercentStart, uPercentSpan);
1677	if (RT_SUCCESS(rc))
1678	{
1679	/* So far the image is opened in read/write mode. Make sure the
1680	* image is opened in read-only mode if the caller requested that. */
1681	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1682	{
1683	qedFreeImage(pImage, false);
1684	rc = qedOpenImage(pImage, uOpenFlags);
1685	if (RT_FAILURE(rc))
1686	{
1687	RTMemFree(pImage);
1688	goto out;
1689	}
1690	}
1691	*ppBackendData = pImage;
1692	}
1693	else
1694	RTMemFree(pImage);
1695
1696	out:
1697	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1698	return rc;
1699	}
1700
1701	/** @copydoc VDIMAGEBACKEND::pfnRename */
1702	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1703	{
1704	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1705	int rc = VINF_SUCCESS;
1706	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1707
1708	/* Check arguments. */
1709	if ( !pImage
1710	\|\| !pszFilename
1711	\|\| !*pszFilename)
1712	{
1713	rc = VERR_INVALID_PARAMETER;
1714	goto out;
1715	}
1716
1717	/* Close the image. */
1718	rc = qedFreeImage(pImage, false);
1719	if (RT_FAILURE(rc))
1720	goto out;
1721
1722	/* Rename the file. */
1723	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1724	if (RT_FAILURE(rc))
1725	{
1726	/* The move failed, try to reopen the original image. */
1727	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1728	if (RT_FAILURE(rc2))
1729	rc = rc2;
1730
1731	goto out;
1732	}
1733
1734	/* Update pImage with the new information. */
1735	pImage->pszFilename = pszFilename;
1736
1737	/* Open the old image with new name. */
1738	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1739	if (RT_FAILURE(rc))
1740	goto out;
1741
1742	out:
1743	LogFlowFunc(("returns %Rrc\n", rc));
1744	return rc;
1745	}
1746
1747	/** @copydoc VDIMAGEBACKEND::pfnClose */
1748	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1749	{
1750	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1751	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1752	int rc;
1753
1754	rc = qedFreeImage(pImage, fDelete);
1755	RTMemFree(pImage);
1756
1757	LogFlowFunc(("returns %Rrc\n", rc));
1758	return rc;
1759	}
1760
1761	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1762	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1763	{
1764	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1765	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1766	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1767	uint32_t offCluster = 0;
1768	uint32_t idxL1 = 0;
1769	uint32_t idxL2 = 0;
1770	uint64_t offFile = 0;
1771	int rc;
1772
1773	AssertPtr(pImage);
1774	Assert(uOffset % 512 == 0);
1775	Assert(cbToRead % 512 == 0);
1776
1777	if (!VALID_PTR(pIoCtx) \|\| !cbToRead)
1778	{
1779	rc = VERR_INVALID_PARAMETER;
1780	goto out;
1781	}
1782
1783	if ( uOffset + cbToRead > pImage->cbSize
1784	\|\| cbToRead == 0)
1785	{
1786	rc = VERR_INVALID_PARAMETER;
1787	goto out;
1788	}
1789
1790	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1791
1792	/* Clip read size to remain in the cluster. */
1793	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1794
1795	/* Get offset in image. */
1796	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1797	if (RT_SUCCESS(rc))
1798	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1799	pIoCtx, cbToRead);
1800
1801	if ( ( RT_SUCCESS(rc)
1802	\|\| rc == VERR_VD_BLOCK_FREE
1803	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1804	&& pcbActuallyRead)
1805	*pcbActuallyRead = cbToRead;
1806
1807	out:
1808	LogFlowFunc(("returns %Rrc\n", rc));
1809	return rc;
1810	}
1811
1812	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1813	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1814	size_t *pcbPostRead, unsigned fWrite)
1815	{
1816	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1817	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1818	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1819	uint32_t offCluster = 0;
1820	uint32_t idxL1 = 0;
1821	uint32_t idxL2 = 0;
1822	uint64_t offImage = 0;
1823	int rc = VINF_SUCCESS;
1824
1825	AssertPtr(pImage);
1826	Assert(!(uOffset % 512));
1827	Assert(!(cbToWrite % 512));
1828
1829	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
1830	{
1831	rc = VERR_VD_IMAGE_READ_ONLY;
1832	goto out;
1833	}
1834
1835	if (!VALID_PTR(pIoCtx) \|\| !cbToWrite)
1836	{
1837	rc = VERR_INVALID_PARAMETER;
1838	goto out;
1839	}
1840
1841	if ( uOffset + cbToWrite > pImage->cbSize
1842	\|\| cbToWrite == 0)
1843	{
1844	rc = VERR_INVALID_PARAMETER;
1845	goto out;
1846	}
1847
1848	/* Convert offset to L1, L2 index and cluster offset. */
1849	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1850
1851	/* Clip write size to remain in the cluster. */
1852	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1853	Assert(!(cbToWrite % 512));
1854
1855	/* Get offset in image. */
1856	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1857	if (RT_SUCCESS(rc))
1858	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1859	offImage, pIoCtx, cbToWrite, NULL, NULL);
1860	else if (rc == VERR_VD_BLOCK_FREE)
1861	{
1862	if ( cbToWrite == pImage->cbCluster
1863	&& !(fWrite & VD_WRITE_NO_ALLOC))
1864	{
1865	PQEDL2CACHEENTRY pL2Entry = NULL;
1866
1867	/* Full cluster write to previously unallocated cluster.
1868	* Allocate cluster and write data. */
1869	Assert(!offCluster);
1870
1871	do
1872	{
1873	/* Check if we have to allocate a new cluster for L2 tables. */
1874	if (!pImage->paL1Table[idxL1])
1875	{
1876	uint64_t offL2Tbl;
1877	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1878
1879	/* Allocate new async cluster allocation state. */
1880	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1881	if (RT_UNLIKELY(!pL2ClusterAlloc))
1882	{
1883	rc = VERR_NO_MEMORY;
1884	break;
1885	}
1886
1887	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1888	if (!pL2Entry)
1889	{
1890	rc = VERR_NO_MEMORY;
1891	RTMemFree(pL2ClusterAlloc);
1892	break;
1893	}
1894
1895	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1896	pL2Entry->offL2Tbl = offL2Tbl;
1897	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1898
1899	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1900	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1901	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1902	pL2ClusterAlloc->idxL1 = idxL1;
1903	pL2ClusterAlloc->idxL2 = idxL2;
1904	pL2ClusterAlloc->cbToWrite = cbToWrite;
1905	pL2ClusterAlloc->pL2Entry = pL2Entry;
1906
1907	/*
1908	* Write the L2 table first and link to the L1 table afterwards.
1909	* If something unexpected happens the worst case which can happen
1910	* is a leak of some clusters.
1911	*/
1912	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1913	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1914	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1915	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1916	break;
1917	else if (RT_FAILURE(rc))
1918	{
1919	RTMemFree(pL2ClusterAlloc);
1920	qedL2TblCacheEntryFree(pImage, pL2Entry);
1921	break;
1922	}
1923
1924	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1925	}
1926	else
1927	{
1928	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1929	&pL2Entry);
1930
1931	if (RT_SUCCESS(rc))
1932	{
1933	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1934
1935	/* Allocate new async cluster allocation state. */
1936	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1937	if (RT_UNLIKELY(!pDataClusterAlloc))
1938	{
1939	rc = VERR_NO_MEMORY;
1940	break;
1941	}
1942
1943	/* Allocate new cluster for the data. */
1944	uint64_t offData = qedClusterAllocate(pImage, 1);
1945
1946	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1947	pDataClusterAlloc->cbImageOld = offData;
1948	pDataClusterAlloc->offClusterNew = offData;
1949	pDataClusterAlloc->idxL1 = idxL1;
1950	pDataClusterAlloc->idxL2 = idxL2;
1951	pDataClusterAlloc->cbToWrite = cbToWrite;
1952	pDataClusterAlloc->pL2Entry = pL2Entry;
1953
1954	/* Write data. */
1955	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1956	offData, pIoCtx, cbToWrite,
1957	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1958	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1959	break;
1960	else if (RT_FAILURE(rc))
1961	{
1962	RTMemFree(pDataClusterAlloc);
1963	break;
1964	}
1965
1966	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1967	}
1968	}
1969
1970	} while (0);
1971
1972	*pcbPreRead = 0;
1973	*pcbPostRead = 0;
1974	}
1975	else
1976	{
1977	/* Trying to do a partial write to an unallocated cluster. Don't do
1978	* anything except letting the upper layer know what to do. */
1979	*pcbPreRead = offCluster;
1980	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
1981	}
1982	}
1983
1984	if (pcbWriteProcess)
1985	*pcbWriteProcess = cbToWrite;
1986
1987
1988	out:
1989	LogFlowFunc(("returns %Rrc\n", rc));
1990	return rc;
1991	}
1992
1993	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
1994	{
1995	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1996	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1997	int rc = VINF_SUCCESS;
1998
1999	Assert(pImage);
2000
2001	if (VALID_PTR(pIoCtx))
2002	rc = qedFlushImageAsync(pImage, pIoCtx);
2003	else
2004	rc = VERR_INVALID_PARAMETER;
2005
2006	LogFlowFunc(("returns %Rrc\n", rc));
2007	return rc;
2008	}
2009
2010	/** @copydoc VDIMAGEBACKEND::pfnGetVersion */
2011	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
2012	{
2013	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2014	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2015
2016	AssertPtr(pImage);
2017
2018	if (pImage)
2019	return 1;
2020	else
2021	return 0;
2022	}
2023
2024	/** @copydoc VDIMAGEBACKEND::pfnGetSectorSize */
2025	static DECLCALLBACK(uint32_t) qedGetSectorSize(void *pBackendData)
2026	{
2027	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2028	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2029	uint32_t cb = 0;
2030
2031	AssertPtr(pImage);
2032
2033	if (pImage && pImage->pStorage)
2034	cb = 512;
2035
2036	LogFlowFunc(("returns %u\n", cb));
2037	return cb;
2038	}
2039
2040	/** @copydoc VDIMAGEBACKEND::pfnGetSize */
2041	static DECLCALLBACK(uint64_t) qedGetSize(void *pBackendData)
2042	{
2043	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2044	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2045	uint64_t cb = 0;
2046
2047	AssertPtr(pImage);
2048
2049	if (pImage && pImage->pStorage)
2050	cb = pImage->cbSize;
2051
2052	LogFlowFunc(("returns %llu\n", cb));
2053	return cb;
2054	}
2055
2056	/** @copydoc VDIMAGEBACKEND::pfnGetFileSize */
2057	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
2058	{
2059	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2060	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2061	uint64_t cb = 0;
2062
2063	AssertPtr(pImage);
2064
2065	if (pImage)
2066	{
2067	uint64_t cbFile;
2068	if (pImage->pStorage)
2069	{
2070	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
2071	if (RT_SUCCESS(rc))
2072	cb += cbFile;
2073	}
2074	}
2075
2076	LogFlowFunc(("returns %lld\n", cb));
2077	return cb;
2078	}
2079
2080	/** @copydoc VDIMAGEBACKEND::pfnGetPCHSGeometry */
2081	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
2082	PVDGEOMETRY pPCHSGeometry)
2083	{
2084	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
2085	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2086	int rc;
2087
2088	AssertPtr(pImage);
2089
2090	if (pImage)
2091	{
2092	if (pImage->PCHSGeometry.cCylinders)
2093	{
2094	*pPCHSGeometry = pImage->PCHSGeometry;
2095	rc = VINF_SUCCESS;
2096	}
2097	else
2098	rc = VERR_VD_GEOMETRY_NOT_SET;
2099	}
2100	else
2101	rc = VERR_VD_NOT_OPENED;
2102
2103	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2104	return rc;
2105	}
2106
2107	/** @copydoc VDIMAGEBACKEND::pfnSetPCHSGeometry */
2108	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
2109	PCVDGEOMETRY pPCHSGeometry)
2110	{
2111	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n", pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2112	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2113	int rc;
2114
2115	AssertPtr(pImage);
2116
2117	if (pImage)
2118	{
2119	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2120	{
2121	rc = VERR_VD_IMAGE_READ_ONLY;
2122	goto out;
2123	}
2124
2125	pImage->PCHSGeometry = *pPCHSGeometry;
2126	rc = VINF_SUCCESS;
2127	}
2128	else
2129	rc = VERR_VD_NOT_OPENED;
2130
2131	out:
2132	LogFlowFunc(("returns %Rrc\n", rc));
2133	return rc;
2134	}
2135
2136	/** @copydoc VDIMAGEBACKEND::pfnGetLCHSGeometry */
2137	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData,
2138	PVDGEOMETRY pLCHSGeometry)
2139	{
2140	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2141	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2142	int rc;
2143
2144	AssertPtr(pImage);
2145
2146	if (pImage)
2147	{
2148	if (pImage->LCHSGeometry.cCylinders)
2149	{
2150	*pLCHSGeometry = pImage->LCHSGeometry;
2151	rc = VINF_SUCCESS;
2152	}
2153	else
2154	rc = VERR_VD_GEOMETRY_NOT_SET;
2155	}
2156	else
2157	rc = VERR_VD_NOT_OPENED;
2158
2159	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2160	return rc;
2161	}
2162
2163	/** @copydoc VDIMAGEBACKEND::pfnSetLCHSGeometry */
2164	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData,
2165	PCVDGEOMETRY pLCHSGeometry)
2166	{
2167	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData, pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2168	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2169	int rc;
2170
2171	AssertPtr(pImage);
2172
2173	if (pImage)
2174	{
2175	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2176	{
2177	rc = VERR_VD_IMAGE_READ_ONLY;
2178	goto out;
2179	}
2180
2181	pImage->LCHSGeometry = *pLCHSGeometry;
2182	rc = VINF_SUCCESS;
2183	}
2184	else
2185	rc = VERR_VD_NOT_OPENED;
2186
2187	out:
2188	LogFlowFunc(("returns %Rrc\n", rc));
2189	return rc;
2190	}
2191
2192	/** @copydoc VDIMAGEBACKEND::pfnGetImageFlags */
2193	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2194	{
2195	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2196	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2197	unsigned uImageFlags;
2198
2199	AssertPtr(pImage);
2200
2201	if (pImage)
2202	uImageFlags = pImage->uImageFlags;
2203	else
2204	uImageFlags = 0;
2205
2206	LogFlowFunc(("returns %#x\n", uImageFlags));
2207	return uImageFlags;
2208	}
2209
2210	/** @copydoc VDIMAGEBACKEND::pfnGetOpenFlags */
2211	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2212	{
2213	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2214	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2215	unsigned uOpenFlags;
2216
2217	AssertPtr(pImage);
2218
2219	if (pImage)
2220	uOpenFlags = pImage->uOpenFlags;
2221	else
2222	uOpenFlags = 0;
2223
2224	LogFlowFunc(("returns %#x\n", uOpenFlags));
2225	return uOpenFlags;
2226	}
2227
2228	/** @copydoc VDIMAGEBACKEND::pfnSetOpenFlags */
2229	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2230	{
2231	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2232	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2233	int rc;
2234
2235	/* Image must be opened and the new flags must be valid. */
2236	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2237	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2238	{
2239	rc = VERR_INVALID_PARAMETER;
2240	goto out;
2241	}
2242
2243	/* Implement this operation via reopening the image. */
2244	rc = qedFreeImage(pImage, false);
2245	if (RT_FAILURE(rc))
2246	goto out;
2247	rc = qedOpenImage(pImage, uOpenFlags);
2248
2249	out:
2250	LogFlowFunc(("returns %Rrc\n", rc));
2251	return rc;
2252	}
2253
2254	/** @copydoc VDIMAGEBACKEND::pfnGetComment */
2255	static DECLCALLBACK(int) qedGetComment(void pBackendData, char pszComment,
2256	size_t cbComment)
2257	{
2258	RT_NOREF2(pszComment, cbComment);
2259	LogFlowFunc(("pBackendData=%#p pszComment=%#p cbComment=%zu\n", pBackendData, pszComment, cbComment));
2260	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2261	int rc;
2262
2263	AssertPtr(pImage);
2264
2265	if (pImage)
2266	rc = VERR_NOT_SUPPORTED;
2267	else
2268	rc = VERR_VD_NOT_OPENED;
2269
2270	LogFlowFunc(("returns %Rrc comment='%s'\n", rc, pszComment));
2271	return rc;
2272	}
2273
2274	/** @copydoc VDIMAGEBACKEND::pfnSetComment */
2275	static DECLCALLBACK(int) qedSetComment(void pBackendData, const char pszComment)
2276	{
2277	RT_NOREF1(pszComment);
2278	LogFlowFunc(("pBackendData=%#p pszComment=\"%s\"\n", pBackendData, pszComment));
2279	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2280	int rc;
2281
2282	AssertPtr(pImage);
2283
2284	if (pImage)
2285	{
2286	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2287	rc = VERR_VD_IMAGE_READ_ONLY;
2288	else
2289	rc = VERR_NOT_SUPPORTED;
2290	}
2291	else
2292	rc = VERR_VD_NOT_OPENED;
2293
2294	LogFlowFunc(("returns %Rrc\n", rc));
2295	return rc;
2296	}
2297
2298	/** @copydoc VDIMAGEBACKEND::pfnGetUuid */
2299	static DECLCALLBACK(int) qedGetUuid(void *pBackendData, PRTUUID pUuid)
2300	{
2301	RT_NOREF1(pUuid);
2302	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2303	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2304	int rc;
2305
2306	AssertPtr(pImage);
2307
2308	if (pImage)
2309	rc = VERR_NOT_SUPPORTED;
2310	else
2311	rc = VERR_VD_NOT_OPENED;
2312
2313	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2314	return rc;
2315	}
2316
2317	/** @copydoc VDIMAGEBACKEND::pfnSetUuid */
2318	static DECLCALLBACK(int) qedSetUuid(void *pBackendData, PCRTUUID pUuid)
2319	{
2320	RT_NOREF1(pUuid);
2321	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2322	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2323	int rc;
2324
2325	LogFlowFunc(("%RTuuid\n", pUuid));
2326	AssertPtr(pImage);
2327
2328	if (pImage)
2329	{
2330	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2331	rc = VERR_NOT_SUPPORTED;
2332	else
2333	rc = VERR_VD_IMAGE_READ_ONLY;
2334	}
2335	else
2336	rc = VERR_VD_NOT_OPENED;
2337
2338	LogFlowFunc(("returns %Rrc\n", rc));
2339	return rc;
2340	}
2341
2342	/** @copydoc VDIMAGEBACKEND::pfnGetModificationUuid */
2343	static DECLCALLBACK(int) qedGetModificationUuid(void *pBackendData, PRTUUID pUuid)
2344	{
2345	RT_NOREF1(pUuid);
2346	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2347	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2348	int rc;
2349
2350	AssertPtr(pImage);
2351
2352	if (pImage)
2353	rc = VERR_NOT_SUPPORTED;
2354	else
2355	rc = VERR_VD_NOT_OPENED;
2356
2357	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2358	return rc;
2359	}
2360
2361	/** @copydoc VDIMAGEBACKEND::pfnSetModificationUuid */
2362	static DECLCALLBACK(int) qedSetModificationUuid(void *pBackendData, PCRTUUID pUuid)
2363	{
2364	RT_NOREF1(pUuid);
2365	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2366	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2367	int rc;
2368
2369	AssertPtr(pImage);
2370
2371	if (pImage)
2372	{
2373	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2374	rc = VERR_NOT_SUPPORTED;
2375	else
2376	rc = VERR_VD_IMAGE_READ_ONLY;
2377	}
2378	else
2379	rc = VERR_VD_NOT_OPENED;
2380
2381	LogFlowFunc(("returns %Rrc\n", rc));
2382	return rc;
2383	}
2384
2385	/** @copydoc VDIMAGEBACKEND::pfnGetParentUuid */
2386	static DECLCALLBACK(int) qedGetParentUuid(void *pBackendData, PRTUUID pUuid)
2387	{
2388	RT_NOREF1(pUuid);
2389	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2390	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2391	int rc;
2392
2393	AssertPtr(pImage);
2394
2395	if (pImage)
2396	rc = VERR_NOT_SUPPORTED;
2397	else
2398	rc = VERR_VD_NOT_OPENED;
2399
2400	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2401	return rc;
2402	}
2403
2404	/** @copydoc VDIMAGEBACKEND::pfnSetParentUuid */
2405	static DECLCALLBACK(int) qedSetParentUuid(void *pBackendData, PCRTUUID pUuid)
2406	{
2407	RT_NOREF1(pUuid);
2408	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2409	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2410	int rc;
2411
2412	AssertPtr(pImage);
2413
2414	if (pImage)
2415	{
2416	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2417	rc = VERR_NOT_SUPPORTED;
2418	else
2419	rc = VERR_VD_IMAGE_READ_ONLY;
2420	}
2421	else
2422	rc = VERR_VD_NOT_OPENED;
2423
2424	LogFlowFunc(("returns %Rrc\n", rc));
2425	return rc;
2426	}
2427
2428	/** @copydoc VDIMAGEBACKEND::pfnGetParentModificationUuid */
2429	static DECLCALLBACK(int) qedGetParentModificationUuid(void *pBackendData, PRTUUID pUuid)
2430	{
2431	RT_NOREF1(pUuid);
2432	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2433	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2434	int rc;
2435
2436	AssertPtr(pImage);
2437
2438	if (pImage)
2439	rc = VERR_NOT_SUPPORTED;
2440	else
2441	rc = VERR_VD_NOT_OPENED;
2442
2443	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2444	return rc;
2445	}
2446
2447	/** @copydoc VDIMAGEBACKEND::pfnSetParentModificationUuid */
2448	static DECLCALLBACK(int) qedSetParentModificationUuid(void *pBackendData, PCRTUUID pUuid)
2449	{
2450	RT_NOREF1(pUuid);
2451	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2452	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2453	int rc;
2454
2455	AssertPtr(pImage);
2456
2457	if (pImage)
2458	{
2459	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2460	rc = VERR_NOT_SUPPORTED;
2461	else
2462	rc = VERR_VD_IMAGE_READ_ONLY;
2463	}
2464	else
2465	rc = VERR_VD_NOT_OPENED;
2466
2467	LogFlowFunc(("returns %Rrc\n", rc));
2468	return rc;
2469	}
2470
2471	/** @copydoc VDIMAGEBACKEND::pfnDump */
2472	static DECLCALLBACK(void) qedDump(void *pBackendData)
2473	{
2474	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2475
2476	AssertPtr(pImage);
2477	if (pImage)
2478	{
2479	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2480	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2481	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2482	pImage->cbSize / 512);
2483	}
2484	}
2485
2486	/** @copydoc VDIMAGEBACKEND::pfnGetParentFilename */
2487	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2488	{
2489	int rc = VINF_SUCCESS;
2490	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2491
2492	AssertPtr(pImage);
2493	if (pImage)
2494	if (pImage->pszBackingFilename)
2495	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2496	else
2497	rc = VERR_NOT_SUPPORTED;
2498	else
2499	rc = VERR_VD_NOT_OPENED;
2500
2501	LogFlowFunc(("returns %Rrc\n", rc));
2502	return rc;
2503	}
2504
2505	/** @copydoc VDIMAGEBACKEND::pfnSetParentFilename */
2506	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2507	{
2508	int rc = VINF_SUCCESS;
2509	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2510
2511	AssertPtr(pImage);
2512	if (pImage)
2513	{
2514	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2515	rc = VERR_VD_IMAGE_READ_ONLY;
2516	else if ( pImage->pszBackingFilename
2517	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2518	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2519	else
2520	{
2521	if (pImage->pszBackingFilename)
2522	RTStrFree(pImage->pszBackingFilename);
2523	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2524	if (!pImage->pszBackingFilename)
2525	rc = VERR_NO_MEMORY;
2526	else
2527	{
2528	if (!pImage->offBackingFilename)
2529	{
2530	/* Allocate new cluster. */
2531	uint64_t offData = qedClusterAllocate(pImage, 1);
2532
2533	Assert((offData & UINT32_MAX) == offData);
2534	pImage->offBackingFilename = (uint32_t)offData;
2535	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2536	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2537	offData + pImage->cbCluster);
2538	}
2539
2540	if (RT_SUCCESS(rc))
2541	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2542	pImage->offBackingFilename,
2543	pImage->pszBackingFilename,
2544	strlen(pImage->pszBackingFilename));
2545	}
2546	}
2547	}
2548	else
2549	rc = VERR_VD_NOT_OPENED;
2550
2551	LogFlowFunc(("returns %Rrc\n", rc));
2552	return rc;
2553	}
2554
2555	/** @copydoc VDIMAGEBACKEND::pfnResize */
2556	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2557	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2558	unsigned uPercentStart, unsigned uPercentSpan,
2559	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2560	PVDINTERFACE pVDIfsOperation)
2561	{
2562	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2563	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2564	int rc = VINF_SUCCESS;
2565
2566	/* Making the image smaller is not supported at the moment. */
2567	if (cbSize < pImage->cbSize)
2568	rc = VERR_NOT_SUPPORTED;
2569	else if (cbSize > pImage->cbSize)
2570	{
2571	/*
2572	* It is enough to just update the size field in the header to complete
2573	* growing. With the default cluster and table sizes the image can be expanded
2574	* to 64TB without overflowing the L1 and L2 tables making block relocation
2575	* superfluous.
2576	* @todo: The rare case where block relocation is still required (non default
2577	* table and/or cluster size or images with more than 64TB) is not
2578	* implemented yet and resizing such an image will fail with an error.
2579	*/
2580	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2581	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2582	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2583	pImage->pszFilename);
2584	else
2585	{
2586	uint64_t cbSizeOld = pImage->cbSize;
2587
2588	pImage->cbSize = cbSize;
2589	rc = qedFlushImage(pImage);
2590	if (RT_FAILURE(rc))
2591	{
2592	pImage->cbSize = cbSizeOld; /* Restore */
2593
2594	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2595	pImage->pszFilename);
2596	}
2597	}
2598	}
2599	/* Same size doesn't change the image at all. */
2600
2601	LogFlowFunc(("returns %Rrc\n", rc));
2602	return rc;
2603	}
2604
2605
2606	const VDIMAGEBACKEND g_QedBackend =
2607	{
2608	/* pszBackendName */
2609	"QED",
2610	/* cbSize */
2611	sizeof(VDIMAGEBACKEND),
2612	/* uBackendCaps */
2613	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2614	/* paFileExtensions */
2615	s_aQedFileExtensions,
2616	/* paConfigInfo */
2617	NULL,
2618	/* pfnCheckIfValid */
2619	qedCheckIfValid,
2620	/* pfnOpen */
2621	qedOpen,
2622	/* pfnCreate */
2623	qedCreate,
2624	/* pfnRename */
2625	qedRename,
2626	/* pfnClose */
2627	qedClose,
2628	/* pfnRead */
2629	qedRead,
2630	/* pfnWrite */
2631	qedWrite,
2632	/* pfnFlush */
2633	qedFlush,
2634	/* pfnDiscard */
2635	NULL,
2636	/* pfnGetVersion */
2637	qedGetVersion,
2638	/* pfnGetSectorSize */
2639	qedGetSectorSize,
2640	/* pfnGetSize */
2641	qedGetSize,
2642	/* pfnGetFileSize */
2643	qedGetFileSize,
2644	/* pfnGetPCHSGeometry */
2645	qedGetPCHSGeometry,
2646	/* pfnSetPCHSGeometry */
2647	qedSetPCHSGeometry,
2648	/* pfnGetLCHSGeometry */
2649	qedGetLCHSGeometry,
2650	/* pfnSetLCHSGeometry */
2651	qedSetLCHSGeometry,
2652	/* pfnGetImageFlags */
2653	qedGetImageFlags,
2654	/* pfnGetOpenFlags */
2655	qedGetOpenFlags,
2656	/* pfnSetOpenFlags */
2657	qedSetOpenFlags,
2658	/* pfnGetComment */
2659	qedGetComment,
2660	/* pfnSetComment */
2661	qedSetComment,
2662	/* pfnGetUuid */
2663	qedGetUuid,
2664	/* pfnSetUuid */
2665	qedSetUuid,
2666	/* pfnGetModificationUuid */
2667	qedGetModificationUuid,
2668	/* pfnSetModificationUuid */
2669	qedSetModificationUuid,
2670	/* pfnGetParentUuid */
2671	qedGetParentUuid,
2672	/* pfnSetParentUuid */
2673	qedSetParentUuid,
2674	/* pfnGetParentModificationUuid */
2675	qedGetParentModificationUuid,
2676	/* pfnSetParentModificationUuid */
2677	qedSetParentModificationUuid,
2678	/* pfnDump */
2679	qedDump,
2680	/* pfnGetTimestamp */
2681	NULL,
2682	/* pfnGetParentTimestamp */
2683	NULL,
2684	/* pfnSetParentTimestamp */
2685	NULL,
2686	/* pfnGetParentFilename */
2687	qedGetParentFilename,
2688	/* pfnSetParentFilename */
2689	qedSetParentFilename,
2690	/* pfnComposeLocation */
2691	genericFileComposeLocation,
2692	/* pfnComposeName */
2693	genericFileComposeName,
2694	/* pfnCompact */
2695	NULL,
2696	/* pfnResize */
2697	qedResize,
2698	/* pfnRepair */
2699	NULL,
2700	/* pfnTraverseMetadata */
2701	NULL
2702	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 63788

Download in other formats: