QED.cpp@ 66486

Last change on this file since 66486 was 66486, checked in by vboxsync, 8 years ago
Storage/VD: Convert all backends to use the region list callbacks, remove the pfnGetSize and pfnGetSectorSize callbacks because they are covered by the region lists
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 90.0 KB

Line
1	/* $Id: QED.cpp 66486 2017-04-10 07:23:59Z 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	/** Pointer to the L2 table we are currently allocating
210	* (can be only one at a time). */
211	PQEDL2CACHEENTRY pL2TblAlloc;
212
213	/** Memory occupied by the L2 table cache. */
214	size_t cbL2Cache;
215	/** The sorted L2 entry list used for searching. */
216	RTLISTNODE ListSearch;
217	/** The LRU L2 entry list used for eviction. */
218	RTLISTNODE ListLru;
219	/** The static region list. */
220	VDREGIONLIST RegionList;
221	} QEDIMAGE, *PQEDIMAGE;
222
223	/**
224	* State of the async cluster allocation.
225	*/
226	typedef enum QEDCLUSTERASYNCALLOCSTATE
227	{
228	/** Invalid. */
229	QEDCLUSTERASYNCALLOCSTATE_INVALID = 0,
230	/** L2 table allocation. */
231	QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC,
232	/** Link L2 table into L1. */
233	QEDCLUSTERASYNCALLOCSTATE_L2_LINK,
234	/** Allocate user data cluster. */
235	QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC,
236	/** Link user data cluster. */
237	QEDCLUSTERASYNCALLOCSTATE_USER_LINK,
238	/** 32bit blowup. */
239	QEDCLUSTERASYNCALLOCSTATE_32BIT_HACK = 0x7fffffff
240	} QEDCLUSTERASYNCALLOCSTATE, *PQEDCLUSTERASYNCALLOCSTATE;
241
242	/**
243	* Data needed to track async cluster allocation.
244	*/
245	typedef struct QEDCLUSTERASYNCALLOC
246	{
247	/** The state of the cluster allocation. */
248	QEDCLUSTERASYNCALLOCSTATE enmAllocState;
249	/** Old image size to rollback in case of an error. */
250	uint64_t cbImageOld;
251	/** L1 index to link if any. */
252	uint32_t idxL1;
253	/** L2 index to link, required in any case. */
254	uint32_t idxL2;
255	/** Start offset of the allocated cluster. */
256	uint64_t offClusterNew;
257	/** L2 cache entry if a L2 table is allocated. */
258	PQEDL2CACHEENTRY pL2Entry;
259	/** Number of bytes to write. */
260	size_t cbToWrite;
261	} QEDCLUSTERASYNCALLOC, *PQEDCLUSTERASYNCALLOC;
262
263
264	/*********************************************************************************************************************************
265	* Static Variables *
266	*********************************************************************************************************************************/
267
268	/** NULL-terminated array of supported file extensions. */
269	static const VDFILEEXTENSION s_aQedFileExtensions[] =
270	{
271	{"qed", VDTYPE_HDD},
272	{NULL, VDTYPE_INVALID}
273	};
274
275
276	/*********************************************************************************************************************************
277	* Internal Functions *
278	*********************************************************************************************************************************/
279
280	/**
281	* Converts the image header to the host endianess and performs basic checks.
282	*
283	* @returns Whether the given header is valid or not.
284	* @param pHeader Pointer to the header to convert.
285	*/
286	static bool qedHdrConvertToHostEndianess(PQedHeader pHeader)
287	{
288	pHeader->u32Magic = RT_LE2H_U32(pHeader->u32Magic);
289	pHeader->u32ClusterSize = RT_LE2H_U32(pHeader->u32ClusterSize);
290	pHeader->u32TableSize = RT_LE2H_U32(pHeader->u32TableSize);
291	pHeader->u32HeaderSize = RT_LE2H_U32(pHeader->u32HeaderSize);
292	pHeader->u64FeatureFlags = RT_LE2H_U64(pHeader->u64FeatureFlags);
293	pHeader->u64CompatFeatureFlags = RT_LE2H_U64(pHeader->u64CompatFeatureFlags);
294	pHeader->u64AutoresetFeatureFlags = RT_LE2H_U64(pHeader->u64AutoresetFeatureFlags);
295	pHeader->u64OffL1Table = RT_LE2H_U64(pHeader->u64OffL1Table);
296	pHeader->u64Size = RT_LE2H_U64(pHeader->u64Size);
297	pHeader->u32OffBackingFilename = RT_LE2H_U32(pHeader->u32OffBackingFilename);
298	pHeader->u32BackingFilenameSize = RT_LE2H_U32(pHeader->u32BackingFilenameSize);
299
300	if (RT_UNLIKELY(pHeader->u32Magic != QED_MAGIC))
301	return false;
302	if (RT_UNLIKELY( pHeader->u32ClusterSize < QED_CLUSTER_SIZE_MIN
303	\|\| pHeader->u32ClusterSize > QED_CLUSTER_SIZE_MAX))
304	return false;
305	if (RT_UNLIKELY( pHeader->u32TableSize < QED_TABLE_SIZE_MIN
306	\|\| pHeader->u32TableSize > QED_TABLE_SIZE_MAX))
307	return false;
308	if (RT_UNLIKELY(pHeader->u64Size % 512 != 0))
309	return false;
310	if (RT_UNLIKELY( pHeader->u64FeatureFlags & QED_FEATURE_BACKING_FILE
311	&& ( pHeader->u32BackingFilenameSize == 0
312	\|\| pHeader->u32BackingFilenameSize == UINT32_MAX)))
313	return false;
314
315	return true;
316	}
317
318	/**
319	* Creates a QED header from the given image state.
320	*
321	* @returns nothing.
322	* @param pImage Image instance data.
323	* @param pHeader Pointer to the header to convert.
324	*/
325	static void qedHdrConvertFromHostEndianess(PQEDIMAGE pImage, PQedHeader pHeader)
326	{
327	pHeader->u32Magic = RT_H2LE_U32(QED_MAGIC);
328	pHeader->u32ClusterSize = RT_H2LE_U32(pImage->cbCluster);
329	pHeader->u32TableSize = RT_H2LE_U32(pImage->cbTable / pImage->cbCluster);
330	pHeader->u32HeaderSize = RT_H2LE_U32(1);
331	pHeader->u64FeatureFlags = RT_H2LE_U64(pImage->pszBackingFilename ? QED_FEATURE_BACKING_FILE : UINT64_C(0));
332	pHeader->u64CompatFeatureFlags = RT_H2LE_U64(UINT64_C(0));
333	pHeader->u64AutoresetFeatureFlags = RT_H2LE_U64(UINT64_C(0));
334	pHeader->u64OffL1Table = RT_H2LE_U64(pImage->offL1Table);
335	pHeader->u64Size = RT_H2LE_U64(pImage->cbSize);
336	pHeader->u32OffBackingFilename = RT_H2LE_U32(pImage->offBackingFilename);
337	pHeader->u32BackingFilenameSize = RT_H2LE_U32(pImage->cbBackingFilename);
338	}
339
340	/**
341	* Convert table entries from little endian to host endianess.
342	*
343	* @returns nothing.
344	* @param paTbl Pointer to the table.
345	* @param cEntries Number of entries in the table.
346	*/
347	static void qedTableConvertToHostEndianess(uint64_t *paTbl, uint32_t cEntries)
348	{
349	while(cEntries-- > 0)
350	{
351	paTbl = RT_LE2H_U64(paTbl);
352	paTbl++;
353	}
354	}
355
356	#if defined(RT_BIG_ENDIAN)
357	/**
358	* Convert table entries from host to little endian format.
359	*
360	* @returns nothing.
361	* @param paTblImg Pointer to the table which will store the little endian table.
362	* @param paTbl The source table to convert.
363	* @param cEntries Number of entries in the table.
364	*/
365	static void qedTableConvertFromHostEndianess(uint64_t paTblImg, uint64_t paTbl,
366	uint32_t cEntries)
367	{
368	while(cEntries-- > 0)
369	{
370	paTblImg = RT_H2LE_U64(paTbl);
371	paTbl++;
372	paTblImg++;
373	}
374	}
375	#endif
376
377	/**
378	* Creates the L2 table cache.
379	*
380	* @returns VBox status code.
381	* @param pImage The image instance data.
382	*/
383	static int qedL2TblCacheCreate(PQEDIMAGE pImage)
384	{
385	pImage->cbL2Cache = 0;
386	RTListInit(&pImage->ListSearch);
387	RTListInit(&pImage->ListLru);
388
389	return VINF_SUCCESS;
390	}
391
392	/**
393	* Destroys the L2 table cache.
394	*
395	* @returns nothing.
396	* @param pImage The image instance data.
397	*/
398	static void qedL2TblCacheDestroy(PQEDIMAGE pImage)
399	{
400	PQEDL2CACHEENTRY pL2Entry;
401	PQEDL2CACHEENTRY pL2Next;
402	RTListForEachSafe(&pImage->ListSearch, pL2Entry, pL2Next, QEDL2CACHEENTRY, NodeSearch)
403	{
404	Assert(!pL2Entry->cRefs);
405
406	RTListNodeRemove(&pL2Entry->NodeSearch);
407	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
408	RTMemFree(pL2Entry);
409	}
410
411	pImage->cbL2Cache = 0;
412	RTListInit(&pImage->ListSearch);
413	RTListInit(&pImage->ListLru);
414	}
415
416	/**
417	* Returns the L2 table matching the given offset or NULL if none could be found.
418	*
419	* @returns Pointer to the L2 table cache entry or NULL.
420	* @param pImage The image instance data.
421	* @param offL2Tbl Offset of the L2 table to search for.
422	*/
423	static PQEDL2CACHEENTRY qedL2TblCacheRetain(PQEDIMAGE pImage, uint64_t offL2Tbl)
424	{
425	if ( pImage->pL2TblAlloc
426	&& pImage->pL2TblAlloc->offL2Tbl == offL2Tbl)
427	{
428	pImage->pL2TblAlloc->cRefs++;
429	return pImage->pL2TblAlloc;
430	}
431
432	PQEDL2CACHEENTRY pL2Entry;
433	RTListForEach(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch)
434	{
435	if (pL2Entry->offL2Tbl == offL2Tbl)
436	break;
437	}
438
439	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
440	{
441	/* Update LRU list. */
442	RTListNodeRemove(&pL2Entry->NodeLru);
443	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
444	pL2Entry->cRefs++;
445	return pL2Entry;
446	}
447	else
448	return NULL;
449	}
450
451	/**
452	* Releases a L2 table cache entry.
453	*
454	* @returns nothing.
455	* @param pL2Entry The L2 cache entry.
456	*/
457	static void qedL2TblCacheEntryRelease(PQEDL2CACHEENTRY pL2Entry)
458	{
459	Assert(pL2Entry->cRefs > 0);
460	pL2Entry->cRefs--;
461	}
462
463	/**
464	* Allocates a new L2 table from the cache evicting old entries if required.
465	*
466	* @returns Pointer to the L2 cache entry or NULL.
467	* @param pImage The image instance data.
468	*/
469	static PQEDL2CACHEENTRY qedL2TblCacheEntryAlloc(PQEDIMAGE pImage)
470	{
471	PQEDL2CACHEENTRY pL2Entry = NULL;
472
473	if (pImage->cbL2Cache + pImage->cbTable <= QED_L2_CACHE_MEMORY_MAX)
474	{
475	/* Add a new entry. */
476	pL2Entry = (PQEDL2CACHEENTRY)RTMemAllocZ(sizeof(QEDL2CACHEENTRY));
477	if (pL2Entry)
478	{
479	pL2Entry->paL2Tbl = (uint64_t *)RTMemPageAllocZ(pImage->cbTable);
480	if (RT_UNLIKELY(!pL2Entry->paL2Tbl))
481	{
482	RTMemFree(pL2Entry);
483	pL2Entry = NULL;
484	}
485	else
486	{
487	pL2Entry->cRefs = 1;
488	pImage->cbL2Cache += pImage->cbTable;
489	}
490	}
491	}
492	else
493	{
494	/* Evict the last not in use entry and use it */
495	Assert(!RTListIsEmpty(&pImage->ListLru));
496
497	RTListForEachReverse(&pImage->ListLru, pL2Entry, QEDL2CACHEENTRY, NodeLru)
498	{
499	if (!pL2Entry->cRefs)
500	break;
501	}
502
503	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
504	{
505	RTListNodeRemove(&pL2Entry->NodeSearch);
506	RTListNodeRemove(&pL2Entry->NodeLru);
507	pL2Entry->offL2Tbl = 0;
508	pL2Entry->cRefs = 1;
509	}
510	else
511	pL2Entry = NULL;
512	}
513
514	return pL2Entry;
515	}
516
517	/**
518	* Frees a L2 table cache entry.
519	*
520	* @returns nothing.
521	* @param pImage The image instance data.
522	* @param pL2Entry The L2 cache entry to free.
523	*/
524	static void qedL2TblCacheEntryFree(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
525	{
526	Assert(!pL2Entry->cRefs);
527	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
528	RTMemFree(pL2Entry);
529
530	pImage->cbL2Cache -= pImage->cbTable;
531	}
532
533	/**
534	* Inserts an entry in the L2 table cache.
535	*
536	* @returns nothing.
537	* @param pImage The image instance data.
538	* @param pL2Entry The L2 cache entry to insert.
539	*/
540	static void qedL2TblCacheEntryInsert(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
541	{
542	Assert(pL2Entry->offL2Tbl > 0);
543
544	/* Insert at the top of the LRU list. */
545	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
546
547	if (RTListIsEmpty(&pImage->ListSearch))
548	{
549	RTListAppend(&pImage->ListSearch, &pL2Entry->NodeSearch);
550	}
551	else
552	{
553	/* Insert into search list. */
554	PQEDL2CACHEENTRY pIt;
555	pIt = RTListGetFirst(&pImage->ListSearch, QEDL2CACHEENTRY, NodeSearch);
556	if (pIt->offL2Tbl > pL2Entry->offL2Tbl)
557	RTListPrepend(&pImage->ListSearch, &pL2Entry->NodeSearch);
558	else
559	{
560	bool fInserted = false;
561
562	RTListForEach(&pImage->ListSearch, pIt, QEDL2CACHEENTRY, NodeSearch)
563	{
564	Assert(pIt->offL2Tbl != pL2Entry->offL2Tbl);
565	if (pIt->offL2Tbl < pL2Entry->offL2Tbl)
566	{
567	RTListNodeInsertAfter(&pIt->NodeSearch, &pL2Entry->NodeSearch);
568	fInserted = true;
569	break;
570	}
571	}
572	Assert(fInserted);
573	}
574	}
575	}
576
577	/**
578	* Fetches the L2 from the given offset trying the LRU cache first and
579	* reading it from the image after a cache miss - version for async I/O.
580	*
581	* @returns VBox status code.
582	* @param pImage Image instance data.
583	* @param pIoCtx The I/O context.
584	* @param offL2Tbl The offset of the L2 table in the image.
585	* @param ppL2Entry Where to store the L2 table on success.
586	*/
587	static int qedL2TblCacheFetchAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
588	uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
589	{
590	int rc = VINF_SUCCESS;
591
592	/* Try to fetch the L2 table from the cache first. */
593	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
594	if (!pL2Entry)
595	{
596	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
597
598	if (pL2Entry)
599	{
600	/* Read from the image. */
601	PVDMETAXFER pMetaXfer;
602
603	pL2Entry->offL2Tbl = offL2Tbl;
604	rc = vdIfIoIntFileReadMeta(pImage->pIfIo, pImage->pStorage,
605	offL2Tbl, pL2Entry->paL2Tbl,
606	pImage->cbTable, pIoCtx,
607	&pMetaXfer, NULL, NULL);
608	if (RT_SUCCESS(rc))
609	{
610	vdIfIoIntMetaXferRelease(pImage->pIfIo, pMetaXfer);
611	#if defined(RT_BIG_ENDIAN)
612	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
613	#endif
614	qedL2TblCacheEntryInsert(pImage, pL2Entry);
615	}
616	else
617	{
618	qedL2TblCacheEntryRelease(pL2Entry);
619	qedL2TblCacheEntryFree(pImage, pL2Entry);
620	}
621	}
622	else
623	rc = VERR_NO_MEMORY;
624	}
625
626	if (RT_SUCCESS(rc))
627	*ppL2Entry = pL2Entry;
628
629	return rc;
630	}
631
632	/**
633	* Return power of 2 or 0 if num error.
634	*
635	* @returns The power of 2 or 0 if the given number is not a power of 2.
636	* @param u32 The number.
637	*/
638	static uint32_t qedGetPowerOfTwo(uint32_t u32)
639	{
640	if (u32 == 0)
641	return 0;
642	uint32_t uPower2 = 0;
643	while ((u32 & 1) == 0)
644	{
645	u32 >>= 1;
646	uPower2++;
647	}
648	return u32 == 1 ? uPower2 : 0;
649	}
650
651	/**
652	* Sets the L1, L2 and offset bitmasks and L1 and L2 bit shift members.
653	*
654	* @returns nothing.
655	* @param pImage The image instance data.
656	*/
657	static void qedTableMasksInit(PQEDIMAGE pImage)
658	{
659	uint32_t cClusterBits, cTableBits;
660
661	cClusterBits = qedGetPowerOfTwo(pImage->cbCluster);
662	cTableBits = qedGetPowerOfTwo(pImage->cTableEntries);
663
664	Assert(cClusterBits + 2 * cTableBits <= 64);
665
666	pImage->fOffsetMask = ((uint64_t)pImage->cbCluster - 1);
667	pImage->fL2Mask = ((uint64_t)pImage->cTableEntries - 1) << cClusterBits;
668	pImage->cL2Shift = cClusterBits;
669	pImage->fL1Mask = ((uint64_t)pImage->cTableEntries - 1) << (cClusterBits + cTableBits);
670	pImage->cL1Shift = cClusterBits + cTableBits;
671	}
672
673	/**
674	* Converts a given logical offset into the
675	*
676	* @returns nothing.
677	* @param pImage The image instance data.
678	* @param off The logical offset to convert.
679	* @param pidxL1 Where to store the index in the L1 table on success.
680	* @param pidxL2 Where to store the index in the L2 table on success.
681	* @param poffCluster Where to store the offset in the cluster on success.
682	*/
683	DECLINLINE(void) qedConvertLogicalOffset(PQEDIMAGE pImage, uint64_t off, uint32_t *pidxL1,
684	uint32_t pidxL2, uint32_t poffCluster)
685	{
686	AssertPtr(pidxL1);
687	AssertPtr(pidxL2);
688	AssertPtr(poffCluster);
689
690	*poffCluster = off & pImage->fOffsetMask;
691	*pidxL1 = (off & pImage->fL1Mask) >> pImage->cL1Shift;
692	*pidxL2 = (off & pImage->fL2Mask) >> pImage->cL2Shift;
693	}
694
695	/**
696	* Converts Cluster size to a byte size.
697	*
698	* @returns Number of bytes derived from the given number of clusters.
699	* @param pImage The image instance data.
700	* @param cClusters The clusters to convert.
701	*/
702	DECLINLINE(uint64_t) qedCluster2Byte(PQEDIMAGE pImage, uint64_t cClusters)
703	{
704	return cClusters * pImage->cbCluster;
705	}
706
707	/**
708	* Converts number of bytes to cluster size rounding to the next cluster.
709	*
710	* @returns Number of bytes derived from the given number of clusters.
711	* @param pImage The image instance data.
712	* @param cb Number of bytes to convert.
713	*/
714	DECLINLINE(uint64_t) qedByte2Cluster(PQEDIMAGE pImage, uint64_t cb)
715	{
716	return cb / pImage->cbCluster + (cb % pImage->cbCluster ? 1 : 0);
717	}
718
719	/**
720	* Allocates a new cluster in the image.
721	*
722	* @returns The start offset of the new cluster in the image.
723	* @param pImage The image instance data.
724	* @param cClusters Number of clusters to allocate.
725	*/
726	DECLINLINE(uint64_t) qedClusterAllocate(PQEDIMAGE pImage, uint32_t cClusters)
727	{
728	uint64_t offCluster;
729
730	offCluster = pImage->cbImage;
731	pImage->cbImage += cClusters*pImage->cbCluster;
732
733	return offCluster;
734	}
735
736	/**
737	* Returns the real image offset for a given cluster or an error if the cluster is not
738	* yet allocated.
739	*
740	* @returns VBox status code.
741	* VERR_VD_BLOCK_FREE if the cluster is not yet allocated.
742	* @param pImage The image instance data.
743	* @param pIoCtx The I/O context.
744	* @param idxL1 The L1 index.
745	* @param idxL2 The L2 index.
746	* @param offCluster Offset inside the cluster.
747	* @param poffImage Where to store the image offset on success;
748	*/
749	static int qedConvertToImageOffset(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
750	uint32_t idxL1, uint32_t idxL2,
751	uint32_t offCluster, uint64_t *poffImage)
752	{
753	int rc = VERR_VD_BLOCK_FREE;
754
755	AssertReturn(idxL1 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
756	AssertReturn(idxL2 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
757
758	if (pImage->paL1Table[idxL1])
759	{
760	PQEDL2CACHEENTRY pL2Entry;
761
762	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
763	&pL2Entry);
764	if (RT_SUCCESS(rc))
765	{
766	/* Get real file offset. */
767	if (pL2Entry->paL2Tbl[idxL2])
768	*poffImage = pL2Entry->paL2Tbl[idxL2] + offCluster;
769	else
770	rc = VERR_VD_BLOCK_FREE;
771
772	qedL2TblCacheEntryRelease(pL2Entry);
773	}
774	}
775
776	return rc;
777	}
778
779	/**
780	* Write the given table to image converting to the image endianess if required.
781	*
782	* @returns VBox status code.
783	* @param pImage The image instance data.
784	* @param pIoCtx The I/O context.
785	* @param offTbl The offset the table should be written to.
786	* @param paTbl The table to write.
787	* @param pfnComplete Callback called when the write completes.
788	* @param pvUser Opaque user data to pass in the completion callback.
789	*/
790	static int qedTblWrite(PQEDIMAGE pImage, PVDIOCTX pIoCtx, uint64_t offTbl, uint64_t *paTbl,
791	PFNVDXFERCOMPLETED pfnComplete, void *pvUser)
792	{
793	int rc = VINF_SUCCESS;
794
795	#if defined(RT_BIG_ENDIAN)
796	uint64_t paTblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
797	if (paTblImg)
798	{
799	qedTableConvertFromHostEndianess(paTblImg, paTbl,
800	pImage->cTableEntries);
801	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
802	offTbl, paTblImg, pImage->cbTable,
803	pIoCtx, pfnComplete, pvUser);
804	RTMemFree(paTblImg);
805	}
806	else
807	rc = VERR_NO_MEMORY;
808	#else
809	/* Write table directly. */
810	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
811	offTbl, paTbl, pImage->cbTable, pIoCtx,
812	pfnComplete, pvUser);
813	#endif
814
815	return rc;
816	}
817
818	/**
819	* Internal. Flush image data to disk.
820	*/
821	static int qedFlushImage(PQEDIMAGE pImage)
822	{
823	int rc = VINF_SUCCESS;
824
825	if ( pImage->pStorage
826	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
827	{
828	QedHeader Header;
829
830	Assert(!(pImage->cbTable % pImage->cbCluster));
831	#if defined(RT_BIG_ENDIAN)
832	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
833	if (paL1TblImg)
834	{
835	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
836	pImage->cTableEntries);
837	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
838	pImage->offL1Table, paL1TblImg,
839	pImage->cbTable);
840	RTMemFree(paL1TblImg);
841	}
842	else
843	rc = VERR_NO_MEMORY;
844	#else
845	/* Write L1 table directly. */
846	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, pImage->offL1Table,
847	pImage->paL1Table, pImage->cbTable);
848	#endif
849	if (RT_SUCCESS(rc))
850	{
851	/* Write header. */
852	qedHdrConvertFromHostEndianess(pImage, &Header);
853	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, 0, &Header,
854	sizeof(Header));
855	if (RT_SUCCESS(rc))
856	rc = vdIfIoIntFileFlushSync(pImage->pIfIo, pImage->pStorage);
857	}
858	}
859
860	return rc;
861	}
862
863	/**
864	* Checks whether the given cluster offset is valid.
865	*
866	* @returns Whether the given cluster offset is valid.
867	* @param offCluster The table offset to check.
868	* @param cbFile The real file size of the image.
869	* @param cbCluster The cluster size in bytes.
870	*/
871	DECLINLINE(bool) qedIsClusterOffsetValid(uint64_t offCluster, uint64_t cbFile, size_t cbCluster)
872	{
873	return (offCluster <= cbFile - cbCluster)
874	&& !(offCluster & (cbCluster - 1));
875	}
876
877	/**
878	* Checks whether the given table offset is valid.
879	*
880	* @returns Whether the given table offset is valid.
881	* @param offTbl The table offset to check.
882	* @param cbFile The real file size of the image.
883	* @param cbTable The table size in bytes.
884	* @param cbCluster The cluster size in bytes.
885	*/
886	DECLINLINE(bool) qedIsTblOffsetValid(uint64_t offTbl, uint64_t cbFile, size_t cbTable, size_t cbCluster)
887	{
888	return (offTbl <= cbFile - cbTable)
889	&& !(offTbl & (cbCluster - 1));
890	}
891
892	/**
893	* Sets the specified range in the cluster bitmap checking whether any of the clusters is already
894	* used before.
895	*
896	* @returns Whether the range was clear and is set now.
897	* @param pvClusterBitmap The cluster bitmap to use.
898	* @param offClusterStart The first cluster to check and set.
899	* @param offClusterEnd The first cluster to not check and set anymore.
900	*/
901	static bool qedClusterBitmapCheckAndSet(void *pvClusterBitmap, uint32_t offClusterStart, uint32_t offClusterEnd)
902	{
903	for (uint32_t offCluster = offClusterStart; offCluster < offClusterEnd; offCluster++)
904	if (ASMBitTest(pvClusterBitmap, offCluster))
905	return false;
906
907	ASMBitSetRange(pvClusterBitmap, offClusterStart, offClusterEnd);
908	return true;
909	}
910
911	/**
912	* Checks the given image for consistency, usually called when the
913	* QED_FEATURE_NEED_CHECK bit is set.
914	*
915	* @returns VBox status code.
916	* @retval VINF_SUCCESS when the image can be accessed.
917	* @param pImage The image instance data.
918	* @param pHeader The header to use for checking.
919	*
920	* @note It is not required that the image state is fully initialized Only
921	* The I/O interface and storage handle need to be valid.
922	* @note The header must be converted to the host CPU endian format already
923	* and should be validated already.
924	*/
925	static int qedCheckImage(PQEDIMAGE pImage, PQedHeader pHeader)
926	{
927	uint64_t cbFile;
928	uint32_t cbTable;
929	uint32_t cTableEntries;
930	uint64_t *paL1Tbl = NULL;
931	uint64_t *paL2Tbl = NULL;
932	void *pvClusterBitmap = NULL;
933	uint32_t offClusterStart;
934	int rc = VINF_SUCCESS;
935
936	pImage->cbCluster = pHeader->u32ClusterSize;
937	cbTable = pHeader->u32TableSize * pHeader->u32ClusterSize;
938	cTableEntries = cbTable / sizeof(uint64_t);
939
940	do
941	{
942	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
943	if (RT_FAILURE(rc))
944	{
945	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
946	N_("Qed: Querying the file size of image '%s' failed"),
947	pImage->pszFilename);
948	break;
949	}
950
951	/* Allocate L1 table. */
952	paL1Tbl = (uint64_t *)RTMemAllocZ(cbTable);
953	if (!paL1Tbl)
954	{
955	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
956	N_("Qed: Allocating memory for the L1 table for image '%s' failed"),
957	pImage->pszFilename);
958	break;
959	}
960
961	paL2Tbl = (uint64_t *)RTMemAllocZ(cbTable);
962	if (!paL2Tbl)
963	{
964	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
965	N_("Qed: Allocating memory for the L2 table for image '%s' failed"),
966	pImage->pszFilename);
967	break;
968	}
969
970	pvClusterBitmap = RTMemAllocZ(cbFile / pHeader->u32ClusterSize / 8);
971	if (!pvClusterBitmap)
972	{
973	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
974	N_("Qed: Allocating memory for the cluster bitmap for image '%s' failed"),
975	pImage->pszFilename);
976	break;
977	}
978
979	/* Validate L1 table offset. */
980	if (!qedIsTblOffsetValid(pHeader->u64OffL1Table, cbFile, cbTable, pHeader->u32ClusterSize))
981	{
982	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
983	N_("Qed: L1 table offset of image '%s' is corrupt (%llu)"),
984	pImage->pszFilename, pHeader->u64OffL1Table);
985	break;
986	}
987
988	/* Read L1 table. */
989	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
990	pHeader->u64OffL1Table, paL1Tbl, cbTable);
991	if (RT_FAILURE(rc))
992	{
993	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
994	N_("Qed: Reading the L1 table from image '%s' failed"),
995	pImage->pszFilename);
996	break;
997	}
998
999	/* Mark the L1 table in cluster bitmap. */
1000	ASMBitSet(pvClusterBitmap, 0); /* Header is always in cluster 0. */
1001	offClusterStart = qedByte2Cluster(pImage, pHeader->u64OffL1Table);
1002	bool fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1003	Assert(fSet);
1004
1005	/* Scan the L1 and L2 tables for invalid entries. */
1006	qedTableConvertToHostEndianess(paL1Tbl, cTableEntries);
1007
1008	for (unsigned iL1 = 0; iL1 < cTableEntries; iL1++)
1009	{
1010	if (!paL1Tbl[iL1])
1011	continue; /* Skip unallocated clusters. */
1012
1013	if (!qedIsTblOffsetValid(paL1Tbl[iL1], cbFile, cbTable, pHeader->u32ClusterSize))
1014	{
1015	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1016	N_("Qed: Entry %d of the L1 table from image '%s' is invalid (%llu)"),
1017	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1018	break;
1019	}
1020
1021	/* Now check that the clusters are not allocated already. */
1022	offClusterStart = qedByte2Cluster(pImage, paL1Tbl[iL1]);
1023	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1024	if (!fSet)
1025	{
1026	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1027	N_("Qed: Entry %d of the L1 table from image '%s' points to a already used cluster (%llu)"),
1028	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1029	break;
1030	}
1031
1032	/* Read the linked L2 table and check it. */
1033	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1034	paL1Tbl[iL1], paL2Tbl, cbTable);
1035	if (RT_FAILURE(rc))
1036	{
1037	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1038	N_("Qed: Reading the L2 table from image '%s' failed"),
1039	pImage->pszFilename);
1040	break;
1041	}
1042
1043	/* Check all L2 entries. */
1044	for (unsigned iL2 = 0; iL2 < cTableEntries; iL2++)
1045	{
1046	if (paL2Tbl[iL2])
1047	continue; /* Skip unallocated clusters. */
1048
1049	if (!qedIsClusterOffsetValid(paL2Tbl[iL2], cbFile, pHeader->u32ClusterSize))
1050	{
1051	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1052	N_("Qed: Entry %d of the L2 table from image '%s' is invalid (%llu)"),
1053	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1054	break;
1055	}
1056
1057	/* Now check that the clusters are not allocated already. */
1058	offClusterStart = qedByte2Cluster(pImage, paL2Tbl[iL2]);
1059	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + 1);
1060	if (!fSet)
1061	{
1062	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1063	N_("Qed: Entry %d of the L2 table from image '%s' points to a already used cluster (%llu)"),
1064	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1065	break;
1066	}
1067	}
1068	}
1069	} while(0);
1070
1071	if (paL1Tbl)
1072	RTMemFree(paL1Tbl);
1073	if (paL2Tbl)
1074	RTMemFree(paL2Tbl);
1075	if (pvClusterBitmap)
1076	RTMemFree(pvClusterBitmap);
1077
1078	return rc;
1079	}
1080
1081	/**
1082	* Internal. Free all allocated space for representing an image except pImage,
1083	* and optionally delete the image from disk.
1084	*/
1085	static int qedFreeImage(PQEDIMAGE pImage, bool fDelete)
1086	{
1087	int rc = VINF_SUCCESS;
1088
1089	/* Freeing a never allocated image (e.g. because the open failed) is
1090	* not signalled as an error. After all nothing bad happens. */
1091	if (pImage)
1092	{
1093	if (pImage->pStorage)
1094	{
1095	/* No point updating the file that is deleted anyway. */
1096	if (!fDelete)
1097	qedFlushImage(pImage);
1098
1099	rc = vdIfIoIntFileClose(pImage->pIfIo, pImage->pStorage);
1100	pImage->pStorage = NULL;
1101	}
1102
1103	if (pImage->paL1Table)
1104	RTMemFree(pImage->paL1Table);
1105
1106	if (pImage->pszBackingFilename)
1107	{
1108	RTMemFree(pImage->pszBackingFilename);
1109	pImage->pszBackingFilename = NULL;
1110	}
1111
1112	qedL2TblCacheDestroy(pImage);
1113
1114	if (fDelete && pImage->pszFilename)
1115	vdIfIoIntFileDelete(pImage->pIfIo, pImage->pszFilename);
1116	}
1117
1118	LogFlowFunc(("returns %Rrc\n", rc));
1119	return rc;
1120	}
1121
1122	/**
1123	* Internal: Open an image, constructing all necessary data structures.
1124	*/
1125	static int qedOpenImage(PQEDIMAGE pImage, unsigned uOpenFlags)
1126	{
1127	pImage->uOpenFlags = uOpenFlags;
1128
1129	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1130	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1131	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1132
1133	/*
1134	* Create the L2 cache before opening the image so we can call qedFreeImage()
1135	* even if opening the image file fails.
1136	*/
1137	int rc = qedL2TblCacheCreate(pImage);
1138	if (RT_SUCCESS(rc))
1139	{
1140	/* Open the image. */
1141	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename,
1142	VDOpenFlagsToFileOpenFlags(uOpenFlags,
1143	false /* fCreate */),
1144	&pImage->pStorage);
1145	if (RT_SUCCESS(rc))
1146	{
1147	uint64_t cbFile;
1148	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1149	if ( RT_SUCCESS(rc)
1150	&& cbFile > sizeof(QedHeader))
1151	{
1152	QedHeader Header;
1153
1154	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, 0, &Header, sizeof(Header));
1155	if ( RT_SUCCESS(rc)
1156	&& qedHdrConvertToHostEndianess(&Header))
1157	{
1158	if ( !(Header.u64FeatureFlags & ~QED_FEATURE_MASK)
1159	&& !(Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE_NO_PROBE))
1160	{
1161	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1162	{
1163	/* Image needs checking. */
1164	if (!(uOpenFlags & VD_OPEN_FLAGS_READONLY))
1165	rc = qedCheckImage(pImage, &Header);
1166	else
1167	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1168	N_("Qed: Image '%s' needs checking but is opened readonly"),
1169	pImage->pszFilename);
1170	}
1171
1172	if ( RT_SUCCESS(rc)
1173	&& (Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE))
1174	{
1175	/* Load backing filename from image. */
1176	pImage->pszBackingFilename = (char )RTMemAllocZ(Header.u32BackingFilenameSize + 1); / +1 for \0 terminator. */
1177	if (pImage->pszBackingFilename)
1178	{
1179	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1180	pImage->offBackingFilename = Header.u32OffBackingFilename;
1181	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1182	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1183	Header.u32BackingFilenameSize);
1184	}
1185	else
1186	rc = VERR_NO_MEMORY;
1187	}
1188
1189	if (RT_SUCCESS(rc))
1190	{
1191	pImage->cbImage = cbFile;
1192	pImage->cbCluster = Header.u32ClusterSize;
1193	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1194	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1195	pImage->offL1Table = Header.u64OffL1Table;
1196	pImage->cbSize = Header.u64Size;
1197	qedTableMasksInit(pImage);
1198
1199	/* Allocate L1 table. */
1200	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1201	if (pImage->paL1Table)
1202	{
1203	/* Read from the image. */
1204	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1205	pImage->offL1Table, pImage->paL1Table,
1206	pImage->cbTable);
1207	if (RT_SUCCESS(rc))
1208	{
1209	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1210
1211	/* If the consistency check succeeded, clear the flag by flushing the image. */
1212	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1213	rc = qedFlushImage(pImage);
1214	}
1215	else
1216	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1217	N_("Qed: Reading the L1 table for image '%s' failed"),
1218	pImage->pszFilename);
1219	}
1220	else
1221	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1222	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1223	pImage->pszFilename);
1224	}
1225	}
1226	else
1227	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1228	N_("Qed: The image '%s' makes use of unsupported features"),
1229	pImage->pszFilename);
1230	}
1231	else if (RT_SUCCESS(rc))
1232	rc = VERR_VD_GEN_INVALID_HEADER;
1233	}
1234	else if (RT_SUCCESS(rc))
1235	rc = VERR_VD_GEN_INVALID_HEADER;
1236	}
1237	/* else: Do NOT signal an appropriate error here, as the VD layer has the
1238	* choice of retrying the open if it failed. */
1239	}
1240	else
1241	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1242	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1243	pImage->pszFilename);
1244
1245	if (RT_SUCCESS(rc))
1246	{
1247	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1248	pImage->RegionList.fFlags = 0;
1249	pImage->RegionList.cRegions = 1;
1250
1251	pRegion->offRegion = 0; /* Disk start. */
1252	pRegion->cbBlock = 512;
1253	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1254	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1255	pRegion->cbData = 512;
1256	pRegion->cbMetadata = 0;
1257	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1258	}
1259	else
1260	qedFreeImage(pImage, false);
1261	return rc;
1262	}
1263
1264	/**
1265	* Internal: Create a qed image.
1266	*/
1267	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1268	unsigned uImageFlags, const char *pszComment,
1269	PCVDGEOMETRY pPCHSGeometry,
1270	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1271	PVDINTERFACEPROGRESS pIfProgress,
1272	unsigned uPercentStart, unsigned uPercentSpan)
1273	{
1274	RT_NOREF1(pszComment);
1275	int rc;
1276
1277	if (!(uImageFlags & VD_IMAGE_FLAGS_FIXED))
1278	{
1279	rc = qedL2TblCacheCreate(pImage);
1280	if (RT_SUCCESS(rc))
1281	{
1282	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1283	pImage->uImageFlags = uImageFlags;
1284	pImage->PCHSGeometry = *pPCHSGeometry;
1285	pImage->LCHSGeometry = *pLCHSGeometry;
1286
1287	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1288	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1289	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1290
1291	/* Create image file. */
1292	uint32_t fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1293	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1294	if (RT_SUCCESS(rc))
1295	{
1296	/* Init image state. */
1297	pImage->cbSize = cbSize;
1298	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1299	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1300	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1301	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1302	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1303	pImage->cbBackingFilename = 0;
1304	pImage->offBackingFilename = 0;
1305	qedTableMasksInit(pImage);
1306
1307	/* Init L1 table. */
1308	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1309	if (RT_LIKELY(pImage->paL1Table))
1310	{
1311	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan * 98 / 100);
1312	rc = qedFlushImage(pImage);
1313	}
1314	else
1315	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1316	pImage->pszFilename);
1317	}
1318	else
1319	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1320	}
1321	else
1322	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1323	pImage->pszFilename);
1324	}
1325	else
1326	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1327
1328	if (RT_SUCCESS(rc))
1329	{
1330	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1331	pImage->RegionList.fFlags = 0;
1332	pImage->RegionList.cRegions = 1;
1333
1334	pRegion->offRegion = 0; /* Disk start. */
1335	pRegion->cbBlock = 512;
1336	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1337	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1338	pRegion->cbData = 512;
1339	pRegion->cbMetadata = 0;
1340	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1341
1342	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan);
1343	}
1344	else
1345	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1346
1347	return rc;
1348	}
1349
1350	/**
1351	* Rollback anything done during async cluster allocation.
1352	*
1353	* @returns VBox status code.
1354	* @param pImage The image instance data.
1355	* @param pIoCtx The I/O context.
1356	* @param pClusterAlloc The cluster allocation to rollback.
1357	*/
1358	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1359	{
1360	RT_NOREF1(pIoCtx);
1361	int rc = VINF_SUCCESS;
1362
1363	switch (pClusterAlloc->enmAllocState)
1364	{
1365	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1366	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1367	{
1368	/* Revert the L1 table entry */
1369	pImage->paL1Table[pClusterAlloc->idxL1] = 0;
1370	pImage->pL2TblAlloc = NULL;
1371
1372	/* Assumption right now is that the L1 table is not modified on storage if the link fails. */
1373	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1374	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1375	Assert(!pClusterAlloc->pL2Entry->cRefs);
1376	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1377	break;
1378	}
1379	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1380	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1381	{
1382	/* Assumption right now is that the L2 table is not modified if the link fails. */
1383	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = 0;
1384	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1385	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1386	break;
1387	}
1388	default:
1389	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1390	rc = VERR_INVALID_STATE;
1391	}
1392
1393	RTMemFree(pClusterAlloc);
1394	return rc;
1395	}
1396
1397	/**
1398	* Updates the state of the async cluster allocation.
1399	*
1400	* @returns VBox status code.
1401	* @param pBackendData The opaque backend data.
1402	* @param pIoCtx I/O context associated with this request.
1403	* @param pvUser Opaque user data passed during a read/write request.
1404	* @param rcReq Status code for the completed request.
1405	*/
1406	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1407	{
1408	int rc = VINF_SUCCESS;
1409	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1410	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1411
1412	if (RT_FAILURE(rcReq))
1413	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1414
1415	AssertPtr(pClusterAlloc->pL2Entry);
1416
1417	switch (pClusterAlloc->enmAllocState)
1418	{
1419	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1420	{
1421	/* Update the link in the in memory L1 table now. */
1422	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1423
1424	/* Update the link in the on disk L1 table now. */
1425	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1426	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1427	qedAsyncClusterAllocUpdate, pClusterAlloc);
1428	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1429	break;
1430	else if (RT_FAILURE(rc))
1431	{
1432	/* Rollback. */
1433	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1434	break;
1435	}
1436	}
1437	/* fall thru */
1438	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1439	{
1440	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1441	uint64_t offData = qedClusterAllocate(pImage, 1);
1442
1443	pImage->pL2TblAlloc = NULL;
1444	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1445
1446	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1447	pClusterAlloc->cbImageOld = offData;
1448	pClusterAlloc->offClusterNew = offData;
1449
1450	/* Write data. */
1451	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1452	offData, pIoCtx, pClusterAlloc->cbToWrite,
1453	qedAsyncClusterAllocUpdate, pClusterAlloc);
1454	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1455	break;
1456	else if (RT_FAILURE(rc))
1457	{
1458	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1459	RTMemFree(pClusterAlloc);
1460	break;
1461	}
1462	}
1463	/* fall thru */
1464	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1465	{
1466	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1467	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1468
1469	/* Link L2 table and update it. */
1470	rc = qedTblWrite(pImage, pIoCtx, pImage->paL1Table[pClusterAlloc->idxL1],
1471	pClusterAlloc->pL2Entry->paL2Tbl,
1472	qedAsyncClusterAllocUpdate, pClusterAlloc);
1473	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1474	break;
1475	else if (RT_FAILURE(rc))
1476	{
1477	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1478	RTMemFree(pClusterAlloc);
1479	break;
1480	}
1481	}
1482	/* fall thru */
1483	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1484	{
1485	/* Everything done without errors, signal completion. */
1486	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1487	RTMemFree(pClusterAlloc);
1488	rc = VINF_SUCCESS;
1489	break;
1490	}
1491	default:
1492	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1493	pClusterAlloc->enmAllocState));
1494	}
1495
1496	return rc;
1497	}
1498
1499	/** @copydoc VDIMAGEBACKEND::pfnProbe */
1500	static DECLCALLBACK(int) qedProbe(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1501	PVDINTERFACE pVDIfsImage, VDTYPE *penmType)
1502	{
1503	RT_NOREF1(pVDIfsDisk);
1504	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1505	PVDIOSTORAGE pStorage = NULL;
1506	int rc = VINF_SUCCESS;
1507
1508	/* Get I/O interface. */
1509	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1510	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1511	AssertReturn((VALID_PTR(pszFilename) && *pszFilename), VERR_INVALID_PARAMETER);
1512
1513	/*
1514	* Open the file and read the footer.
1515	*/
1516	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1517	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1518	false /* fCreate */),
1519	&pStorage);
1520	if (RT_SUCCESS(rc))
1521	{
1522	uint64_t cbFile;
1523
1524	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1525	if ( RT_SUCCESS(rc)
1526	&& cbFile > sizeof(QedHeader))
1527	{
1528	QedHeader Header;
1529
1530	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1531	if ( RT_SUCCESS(rc)
1532	&& qedHdrConvertToHostEndianess(&Header))
1533	*penmType = VDTYPE_HDD;
1534	else
1535	rc = VERR_VD_GEN_INVALID_HEADER;
1536	}
1537	else
1538	rc = VERR_VD_GEN_INVALID_HEADER;
1539	}
1540
1541	if (pStorage)
1542	vdIfIoIntFileClose(pIfIo, pStorage);
1543
1544	LogFlowFunc(("returns %Rrc\n", rc));
1545	return rc;
1546	}
1547
1548	/** @copydoc VDIMAGEBACKEND::pfnOpen */
1549	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1550	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1551	VDTYPE enmType, void **ppBackendData)
1552	{
1553	RT_NOREF1(enmType); /*< @todo r=klaus make use of the type info. /
1554
1555	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n",
1556	pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1557	int rc;
1558
1559	/* Check open flags. All valid flags are supported. */
1560	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1561	AssertReturn((VALID_PTR(pszFilename) && *pszFilename), VERR_INVALID_PARAMETER);
1562
1563	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1564	if (RT_LIKELY(pImage))
1565	{
1566	pImage->pszFilename = pszFilename;
1567	pImage->pStorage = NULL;
1568	pImage->pVDIfsDisk = pVDIfsDisk;
1569	pImage->pVDIfsImage = pVDIfsImage;
1570
1571	rc = qedOpenImage(pImage, uOpenFlags);
1572	if (RT_SUCCESS(rc))
1573	*ppBackendData = pImage;
1574	else
1575	RTMemFree(pImage);
1576	}
1577	else
1578	rc = VERR_NO_MEMORY;
1579
1580	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1581	return rc;
1582	}
1583
1584	/** @copydoc VDIMAGEBACKEND::pfnCreate */
1585	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1586	unsigned uImageFlags, const char *pszComment,
1587	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1588	PCRTUUID pUuid, unsigned uOpenFlags,
1589	unsigned uPercentStart, unsigned uPercentSpan,
1590	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1591	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1592	void **ppBackendData)
1593	{
1594	RT_NOREF1(pUuid);
1595	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",
1596	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1597	int rc;
1598
1599	/* Check the VD container type. */
1600	if (enmType != VDTYPE_HDD)
1601	return VERR_VD_INVALID_TYPE;
1602
1603	/* Check open flags. All valid flags are supported. */
1604	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1605	AssertReturn( VALID_PTR(pszFilename)
1606	&& *pszFilename
1607	&& VALID_PTR(pPCHSGeometry)
1608	&& VALID_PTR(pLCHSGeometry), VERR_INVALID_PARAMETER);
1609
1610	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1611	if (RT_LIKELY(pImage))
1612	{
1613	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1614
1615	pImage->pszFilename = pszFilename;
1616	pImage->pStorage = NULL;
1617	pImage->pVDIfsDisk = pVDIfsDisk;
1618	pImage->pVDIfsImage = pVDIfsImage;
1619
1620	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1621	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1622	pIfProgress, uPercentStart, uPercentSpan);
1623	if (RT_SUCCESS(rc))
1624	{
1625	/* So far the image is opened in read/write mode. Make sure the
1626	* image is opened in read-only mode if the caller requested that. */
1627	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1628	{
1629	qedFreeImage(pImage, false);
1630	rc = qedOpenImage(pImage, uOpenFlags);
1631	}
1632
1633	if (RT_SUCCESS(rc))
1634	*ppBackendData = pImage;
1635	}
1636
1637	if (RT_FAILURE(rc))
1638	RTMemFree(pImage);
1639	}
1640	else
1641	rc = VERR_NO_MEMORY;
1642
1643	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1644	return rc;
1645	}
1646
1647	/** @copydoc VDIMAGEBACKEND::pfnRename */
1648	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1649	{
1650	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1651	int rc = VINF_SUCCESS;
1652	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1653
1654	/* Check arguments. */
1655	AssertReturn((pImage && pszFilename && *pszFilename), VERR_INVALID_PARAMETER);
1656
1657	/* Close the image. */
1658	rc = qedFreeImage(pImage, false);
1659	if (RT_SUCCESS(rc))
1660	{
1661	/* Rename the file. */
1662	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1663	if (RT_SUCCESS(rc))
1664	{
1665	/* Update pImage with the new information. */
1666	pImage->pszFilename = pszFilename;
1667
1668	/* Open the old image with new name. */
1669	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1670	}
1671	else
1672	{
1673	/* The move failed, try to reopen the original image. */
1674	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1675	if (RT_FAILURE(rc2))
1676	rc = rc2;
1677	}
1678	}
1679
1680	LogFlowFunc(("returns %Rrc\n", rc));
1681	return rc;
1682	}
1683
1684	/** @copydoc VDIMAGEBACKEND::pfnClose */
1685	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1686	{
1687	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1688	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1689
1690	int rc = qedFreeImage(pImage, fDelete);
1691	RTMemFree(pImage);
1692
1693	LogFlowFunc(("returns %Rrc\n", rc));
1694	return rc;
1695	}
1696
1697	/** @copydoc VDIMAGEBACKEND::pfnRead */
1698	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1699	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1700	{
1701	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1702	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1703	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1704	uint32_t offCluster = 0;
1705	uint32_t idxL1 = 0;
1706	uint32_t idxL2 = 0;
1707	uint64_t offFile = 0;
1708
1709	AssertPtr(pImage);
1710	Assert(uOffset % 512 == 0);
1711	Assert(cbToRead % 512 == 0);
1712	AssertReturn((VALID_PTR(pIoCtx) && cbToRead), VERR_INVALID_PARAMETER);
1713	AssertReturn(uOffset + cbToRead <= pImage->cbSize, VERR_INVALID_PARAMETER);
1714
1715	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1716
1717	/* Clip read size to remain in the cluster. */
1718	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1719
1720	/* Get offset in image. */
1721	int rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1722	if (RT_SUCCESS(rc))
1723	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1724	pIoCtx, cbToRead);
1725
1726	if ( ( RT_SUCCESS(rc)
1727	\|\| rc == VERR_VD_BLOCK_FREE
1728	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1729	&& pcbActuallyRead)
1730	*pcbActuallyRead = cbToRead;
1731
1732	LogFlowFunc(("returns %Rrc\n", rc));
1733	return rc;
1734	}
1735
1736	/** @copydoc VDIMAGEBACKEND::pfnWrite */
1737	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1738	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1739	size_t *pcbPostRead, unsigned fWrite)
1740	{
1741	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1742	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1743	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1744	uint32_t offCluster = 0;
1745	uint32_t idxL1 = 0;
1746	uint32_t idxL2 = 0;
1747	uint64_t offImage = 0;
1748	int rc = VINF_SUCCESS;
1749
1750	AssertPtr(pImage);
1751	Assert(!(uOffset % 512));
1752	Assert(!(cbToWrite % 512));
1753	AssertReturn((VALID_PTR(pIoCtx) && cbToWrite), VERR_INVALID_PARAMETER);
1754	AssertReturn(uOffset + cbToWrite <= pImage->cbSize, VERR_INVALID_PARAMETER);
1755
1756	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1757	{
1758	/* Convert offset to L1, L2 index and cluster offset. */
1759	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1760
1761	/* Clip write size to remain in the cluster. */
1762	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1763	Assert(!(cbToWrite % 512));
1764
1765	/* Get offset in image. */
1766	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1767	if (RT_SUCCESS(rc))
1768	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1769	offImage, pIoCtx, cbToWrite, NULL, NULL);
1770	else if (rc == VERR_VD_BLOCK_FREE)
1771	{
1772	if ( cbToWrite == pImage->cbCluster
1773	&& !(fWrite & VD_WRITE_NO_ALLOC))
1774	{
1775	PQEDL2CACHEENTRY pL2Entry = NULL;
1776
1777	/* Full cluster write to previously unallocated cluster.
1778	* Allocate cluster and write data. */
1779	Assert(!offCluster);
1780
1781	do
1782	{
1783	/* Check if we have to allocate a new cluster for L2 tables. */
1784	if (!pImage->paL1Table[idxL1])
1785	{
1786	uint64_t offL2Tbl;
1787	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1788
1789	/* Allocate new async cluster allocation state. */
1790	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1791	if (RT_UNLIKELY(!pL2ClusterAlloc))
1792	{
1793	rc = VERR_NO_MEMORY;
1794	break;
1795	}
1796
1797	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1798	if (!pL2Entry)
1799	{
1800	rc = VERR_NO_MEMORY;
1801	RTMemFree(pL2ClusterAlloc);
1802	break;
1803	}
1804
1805	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1806	pL2Entry->offL2Tbl = offL2Tbl;
1807	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1808
1809	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1810	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1811	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1812	pL2ClusterAlloc->idxL1 = idxL1;
1813	pL2ClusterAlloc->idxL2 = idxL2;
1814	pL2ClusterAlloc->cbToWrite = cbToWrite;
1815	pL2ClusterAlloc->pL2Entry = pL2Entry;
1816
1817	pImage->pL2TblAlloc = pL2Entry;
1818
1819	LogFlowFunc(("Allocating new L2 table at cluster offset %llu\n", offL2Tbl));
1820
1821	/*
1822	* Write the L2 table first and link to the L1 table afterwards.
1823	* If something unexpected happens the worst case which can happen
1824	* is a leak of some clusters.
1825	*/
1826	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1827	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1828	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1829	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1830	break;
1831	else if (RT_FAILURE(rc))
1832	{
1833	RTMemFree(pL2ClusterAlloc);
1834	qedL2TblCacheEntryFree(pImage, pL2Entry);
1835	break;
1836	}
1837
1838	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1839	}
1840	else
1841	{
1842	LogFlowFunc(("Fetching L2 table at cluster offset %llu\n", pImage->paL1Table[idxL1]));
1843
1844	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1845	&pL2Entry);
1846
1847	if (RT_SUCCESS(rc))
1848	{
1849	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1850
1851	/* Allocate new async cluster allocation state. */
1852	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1853	if (RT_UNLIKELY(!pDataClusterAlloc))
1854	{
1855	rc = VERR_NO_MEMORY;
1856	break;
1857	}
1858
1859	/* Allocate new cluster for the data. */
1860	uint64_t offData = qedClusterAllocate(pImage, 1);
1861
1862	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1863	pDataClusterAlloc->cbImageOld = offData;
1864	pDataClusterAlloc->offClusterNew = offData;
1865	pDataClusterAlloc->idxL1 = idxL1;
1866	pDataClusterAlloc->idxL2 = idxL2;
1867	pDataClusterAlloc->cbToWrite = cbToWrite;
1868	pDataClusterAlloc->pL2Entry = pL2Entry;
1869
1870	/* Write data. */
1871	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1872	offData, pIoCtx, cbToWrite,
1873	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1874	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1875	break;
1876	else if (RT_FAILURE(rc))
1877	{
1878	RTMemFree(pDataClusterAlloc);
1879	break;
1880	}
1881
1882	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1883	}
1884	}
1885
1886	} while (0);
1887
1888	*pcbPreRead = 0;
1889	*pcbPostRead = 0;
1890	}
1891	else
1892	{
1893	/* Trying to do a partial write to an unallocated cluster. Don't do
1894	* anything except letting the upper layer know what to do. */
1895	*pcbPreRead = offCluster;
1896	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
1897	}
1898	}
1899
1900	if (pcbWriteProcess)
1901	*pcbWriteProcess = cbToWrite;
1902	}
1903	else
1904	rc = VERR_VD_IMAGE_READ_ONLY;
1905
1906	LogFlowFunc(("returns %Rrc\n", rc));
1907	return rc;
1908	}
1909
1910	/** @copydoc VDIMAGEBACKEND::pfnFlush */
1911	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
1912	{
1913	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1914	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1915	int rc = VINF_SUCCESS;
1916
1917	AssertPtr(pImage);
1918	AssertPtrReturn(pIoCtx, VERR_INVALID_PARAMETER);
1919
1920	if ( pImage->pStorage
1921	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1922	{
1923	QedHeader Header;
1924
1925	Assert(!(pImage->cbTable % pImage->cbCluster));
1926	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1927	NULL, NULL);
1928	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1929	{
1930	/* Write header. */
1931	qedHdrConvertFromHostEndianess(pImage, &Header);
1932	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1933	0, &Header, sizeof(Header),
1934	pIoCtx, NULL, NULL);
1935	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1936	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
1937	pIoCtx, NULL, NULL);
1938	}
1939	}
1940
1941	LogFlowFunc(("returns %Rrc\n", rc));
1942	return rc;
1943	}
1944
1945	/** @copydoc VDIMAGEBACKEND::pfnGetVersion */
1946	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
1947	{
1948	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1949	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1950
1951	AssertPtrReturn(pImage, 0);
1952
1953	return 1;
1954	}
1955
1956	/** @copydoc VDIMAGEBACKEND::pfnGetFileSize */
1957	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
1958	{
1959	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1960	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1961	uint64_t cb = 0;
1962
1963	AssertPtrReturn(pImage, 0);
1964
1965	uint64_t cbFile;
1966	if (pImage->pStorage)
1967	{
1968	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1969	if (RT_SUCCESS(rc))
1970	cb += cbFile;
1971	}
1972
1973	LogFlowFunc(("returns %lld\n", cb));
1974	return cb;
1975	}
1976
1977	/** @copydoc VDIMAGEBACKEND::pfnGetPCHSGeometry */
1978	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
1979	PVDGEOMETRY pPCHSGeometry)
1980	{
1981	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
1982	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1983	int rc = VINF_SUCCESS;
1984
1985	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
1986
1987	if (pImage->PCHSGeometry.cCylinders)
1988	*pPCHSGeometry = pImage->PCHSGeometry;
1989	else
1990	rc = VERR_VD_GEOMETRY_NOT_SET;
1991
1992	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
1993	return rc;
1994	}
1995
1996	/** @copydoc VDIMAGEBACKEND::pfnSetPCHSGeometry */
1997	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
1998	PCVDGEOMETRY pPCHSGeometry)
1999	{
2000	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n",
2001	pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2002	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2003	int rc = VINF_SUCCESS;
2004
2005	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2006
2007	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2008	rc = VERR_VD_IMAGE_READ_ONLY;
2009	else
2010	pImage->PCHSGeometry = *pPCHSGeometry;
2011
2012	LogFlowFunc(("returns %Rrc\n", rc));
2013	return rc;
2014	}
2015
2016	/** @copydoc VDIMAGEBACKEND::pfnGetLCHSGeometry */
2017	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData, PVDGEOMETRY pLCHSGeometry)
2018	{
2019	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2020	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2021	int rc = VINF_SUCCESS;
2022
2023	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2024
2025	if (pImage->LCHSGeometry.cCylinders)
2026	*pLCHSGeometry = pImage->LCHSGeometry;
2027	else
2028	rc = VERR_VD_GEOMETRY_NOT_SET;
2029
2030	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders,
2031	pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2032	return rc;
2033	}
2034
2035	/** @copydoc VDIMAGEBACKEND::pfnSetLCHSGeometry */
2036	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData, PCVDGEOMETRY pLCHSGeometry)
2037	{
2038	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData,
2039	pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2040	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2041	int rc = VINF_SUCCESS;
2042
2043	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2044
2045	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2046	rc = VERR_VD_IMAGE_READ_ONLY;
2047	else
2048	pImage->LCHSGeometry = *pLCHSGeometry;
2049
2050	LogFlowFunc(("returns %Rrc\n", rc));
2051	return rc;
2052	}
2053
2054	/** @copydoc VDIMAGEBACKEND::pfnQueryRegions */
2055	static DECLCALLBACK(int) qedQueryRegions(void pBackendData, PCVDREGIONLIST ppRegionList)
2056	{
2057	LogFlowFunc(("pBackendData=%#p ppRegionList=%#p\n", pBackendData, ppRegionList));
2058	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2059
2060	AssertPtrReturn(pThis, VERR_VD_NOT_OPENED);
2061
2062	*ppRegionList = &pThis->RegionList;
2063	LogFlowFunc(("returns %Rrc\n", VINF_SUCCESS));
2064	return VINF_SUCCESS;
2065	}
2066
2067	/** @copydoc VDIMAGEBACKEND::pfnRegionListRelease */
2068	static DECLCALLBACK(void) qedRegionListRelease(void *pBackendData, PCVDREGIONLIST pRegionList)
2069	{
2070	RT_NOREF1(pRegionList);
2071	LogFlowFunc(("pBackendData=%#p pRegionList=%#p\n", pBackendData, pRegionList));
2072	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2073	AssertPtr(pThis); RT_NOREF(pThis);
2074
2075	/* Nothing to do here. */
2076	}
2077
2078	/** @copydoc VDIMAGEBACKEND::pfnGetImageFlags */
2079	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2080	{
2081	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2082	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2083
2084	AssertPtrReturn(pImage, 0);
2085
2086	LogFlowFunc(("returns %#x\n", pImage->uImageFlags));
2087	return pImage->uImageFlags;
2088	}
2089
2090	/** @copydoc VDIMAGEBACKEND::pfnGetOpenFlags */
2091	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2092	{
2093	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2094	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2095
2096	AssertPtrReturn(pImage, 0);
2097
2098	LogFlowFunc(("returns %#x\n", pImage->uOpenFlags));
2099	return pImage->uOpenFlags;
2100	}
2101
2102	/** @copydoc VDIMAGEBACKEND::pfnSetOpenFlags */
2103	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2104	{
2105	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2106	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2107	int rc = VINF_SUCCESS;
2108
2109	/* Image must be opened and the new flags must be valid. */
2110	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2111	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SHAREABLE
2112	\| VD_OPEN_FLAGS_SEQUENTIAL \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2113	rc = VERR_INVALID_PARAMETER;
2114	else
2115	{
2116	/* Implement this operation via reopening the image. */
2117	rc = qedFreeImage(pImage, false);
2118	if (RT_SUCCESS(rc))
2119	rc = qedOpenImage(pImage, uOpenFlags);
2120	}
2121
2122	LogFlowFunc(("returns %Rrc\n", rc));
2123	return rc;
2124	}
2125
2126	/** @copydoc VDIMAGEBACKEND::pfnGetComment */
2127	static DECLCALLBACK(int) qedGetComment(void pBackendData, char pszComment,
2128	size_t cbComment)
2129	{
2130	RT_NOREF2(pszComment, cbComment);
2131	LogFlowFunc(("pBackendData=%#p pszComment=%#p cbComment=%zu\n", pBackendData, pszComment, cbComment));
2132	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2133
2134	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2135
2136	LogFlowFunc(("returns %Rrc comment='%s'\n", VERR_NOT_SUPPORTED, pszComment));
2137	return VERR_NOT_SUPPORTED;
2138	}
2139
2140	/** @copydoc VDIMAGEBACKEND::pfnSetComment */
2141	static DECLCALLBACK(int) qedSetComment(void pBackendData, const char pszComment)
2142	{
2143	RT_NOREF1(pszComment);
2144	LogFlowFunc(("pBackendData=%#p pszComment=\"%s\"\n", pBackendData, pszComment));
2145	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2146
2147	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2148
2149	int rc;
2150	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2151	rc = VERR_VD_IMAGE_READ_ONLY;
2152	else
2153	rc = VERR_NOT_SUPPORTED;
2154
2155	LogFlowFunc(("returns %Rrc\n", rc));
2156	return rc;
2157	}
2158
2159	/** @copydoc VDIMAGEBACKEND::pfnGetUuid */
2160	static DECLCALLBACK(int) qedGetUuid(void *pBackendData, PRTUUID pUuid)
2161	{
2162	RT_NOREF1(pUuid);
2163	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2164	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2165
2166	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2167
2168	LogFlowFunc(("returns %Rrc (%RTuuid)\n", VERR_NOT_SUPPORTED, pUuid));
2169	return VERR_NOT_SUPPORTED;
2170	}
2171
2172	/** @copydoc VDIMAGEBACKEND::pfnSetUuid */
2173	static DECLCALLBACK(int) qedSetUuid(void *pBackendData, PCRTUUID pUuid)
2174	{
2175	RT_NOREF1(pUuid);
2176	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2177	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2178
2179	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2180
2181	int rc;
2182	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2183	rc = VERR_VD_IMAGE_READ_ONLY;
2184	else
2185	rc = VERR_NOT_SUPPORTED;
2186
2187	LogFlowFunc(("returns %Rrc\n", rc));
2188	return rc;
2189	}
2190
2191	/** @copydoc VDIMAGEBACKEND::pfnGetModificationUuid */
2192	static DECLCALLBACK(int) qedGetModificationUuid(void *pBackendData, PRTUUID pUuid)
2193	{
2194	RT_NOREF1(pUuid);
2195	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2196	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2197
2198	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2199
2200	LogFlowFunc(("returns %Rrc (%RTuuid)\n", VERR_NOT_SUPPORTED, pUuid));
2201	return VERR_NOT_SUPPORTED;
2202	}
2203
2204	/** @copydoc VDIMAGEBACKEND::pfnSetModificationUuid */
2205	static DECLCALLBACK(int) qedSetModificationUuid(void *pBackendData, PCRTUUID pUuid)
2206	{
2207	RT_NOREF1(pUuid);
2208	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2209	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2210
2211	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2212
2213	int rc;
2214	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2215	rc = VERR_VD_IMAGE_READ_ONLY;
2216	else
2217	rc = VERR_NOT_SUPPORTED;
2218
2219	LogFlowFunc(("returns %Rrc\n", rc));
2220	return rc;
2221	}
2222
2223	/** @copydoc VDIMAGEBACKEND::pfnGetParentUuid */
2224	static DECLCALLBACK(int) qedGetParentUuid(void *pBackendData, PRTUUID pUuid)
2225	{
2226	RT_NOREF1(pUuid);
2227	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2228	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2229
2230	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2231
2232	LogFlowFunc(("returns %Rrc (%RTuuid)\n", VERR_NOT_SUPPORTED, pUuid));
2233	return VERR_NOT_SUPPORTED;
2234	}
2235
2236	/** @copydoc VDIMAGEBACKEND::pfnSetParentUuid */
2237	static DECLCALLBACK(int) qedSetParentUuid(void *pBackendData, PCRTUUID pUuid)
2238	{
2239	RT_NOREF1(pUuid);
2240	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2241	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2242
2243	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2244
2245	int rc;
2246	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2247	rc = VERR_VD_IMAGE_READ_ONLY;
2248	else
2249	rc = VERR_NOT_SUPPORTED;
2250
2251	LogFlowFunc(("returns %Rrc\n", rc));
2252	return rc;
2253	}
2254
2255	/** @copydoc VDIMAGEBACKEND::pfnGetParentModificationUuid */
2256	static DECLCALLBACK(int) qedGetParentModificationUuid(void *pBackendData, PRTUUID pUuid)
2257	{
2258	RT_NOREF1(pUuid);
2259	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2260	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2261
2262	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2263
2264	LogFlowFunc(("returns %Rrc (%RTuuid)\n", VERR_NOT_SUPPORTED, pUuid));
2265	return VERR_NOT_SUPPORTED;
2266	}
2267
2268	/** @copydoc VDIMAGEBACKEND::pfnSetParentModificationUuid */
2269	static DECLCALLBACK(int) qedSetParentModificationUuid(void *pBackendData, PCRTUUID pUuid)
2270	{
2271	RT_NOREF1(pUuid);
2272	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2273	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2274
2275	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2276
2277	int rc;
2278	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2279	rc = VERR_VD_IMAGE_READ_ONLY;
2280	else
2281	rc = VERR_NOT_SUPPORTED;
2282
2283	LogFlowFunc(("returns %Rrc\n", rc));
2284	return rc;
2285	}
2286
2287	/** @copydoc VDIMAGEBACKEND::pfnDump */
2288	static DECLCALLBACK(void) qedDump(void *pBackendData)
2289	{
2290	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2291
2292	AssertPtrReturnVoid(pImage);
2293	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2294	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2295	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2296	pImage->cbSize / 512);
2297	}
2298
2299	/** @copydoc VDIMAGEBACKEND::pfnGetParentFilename */
2300	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2301	{
2302	int rc = VINF_SUCCESS;
2303	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2304
2305	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2306
2307	if (pImage->pszBackingFilename)
2308	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2309	else
2310	rc = VERR_NOT_SUPPORTED;
2311
2312	LogFlowFunc(("returns %Rrc\n", rc));
2313	return rc;
2314	}
2315
2316	/** @copydoc VDIMAGEBACKEND::pfnSetParentFilename */
2317	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2318	{
2319	int rc = VINF_SUCCESS;
2320	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2321
2322	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2323
2324	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2325	rc = VERR_VD_IMAGE_READ_ONLY;
2326	else if ( pImage->pszBackingFilename
2327	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2328	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2329	else
2330	{
2331	if (pImage->pszBackingFilename)
2332	RTStrFree(pImage->pszBackingFilename);
2333	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2334	if (!pImage->pszBackingFilename)
2335	rc = VERR_NO_MEMORY;
2336	else
2337	{
2338	if (!pImage->offBackingFilename)
2339	{
2340	/* Allocate new cluster. */
2341	uint64_t offData = qedClusterAllocate(pImage, 1);
2342
2343	Assert((offData & UINT32_MAX) == offData);
2344	pImage->offBackingFilename = (uint32_t)offData;
2345	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2346	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2347	offData + pImage->cbCluster);
2348	}
2349
2350	if (RT_SUCCESS(rc))
2351	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2352	pImage->offBackingFilename,
2353	pImage->pszBackingFilename,
2354	strlen(pImage->pszBackingFilename));
2355	}
2356	}
2357
2358	LogFlowFunc(("returns %Rrc\n", rc));
2359	return rc;
2360	}
2361
2362	/** @copydoc VDIMAGEBACKEND::pfnResize */
2363	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2364	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2365	unsigned uPercentStart, unsigned uPercentSpan,
2366	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2367	PVDINTERFACE pVDIfsOperation)
2368	{
2369	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2370	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2371	int rc = VINF_SUCCESS;
2372
2373	/* Making the image smaller is not supported at the moment. */
2374	if (cbSize < pImage->cbSize)
2375	rc = VERR_NOT_SUPPORTED;
2376	else if (cbSize > pImage->cbSize)
2377	{
2378	/*
2379	* It is enough to just update the size field in the header to complete
2380	* growing. With the default cluster and table sizes the image can be expanded
2381	* to 64TB without overflowing the L1 and L2 tables making block relocation
2382	* superfluous.
2383	* @todo: The rare case where block relocation is still required (non default
2384	* table and/or cluster size or images with more than 64TB) is not
2385	* implemented yet and resizing such an image will fail with an error.
2386	*/
2387	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2388	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2389	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2390	pImage->pszFilename);
2391	else
2392	{
2393	uint64_t cbSizeOld = pImage->cbSize;
2394
2395	pImage->cbSize = cbSize;
2396	rc = qedFlushImage(pImage);
2397	if (RT_FAILURE(rc))
2398	{
2399	pImage->cbSize = cbSizeOld; /* Restore */
2400
2401	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2402	pImage->pszFilename);
2403	}
2404	}
2405	}
2406	/* Same size doesn't change the image at all. */
2407
2408	LogFlowFunc(("returns %Rrc\n", rc));
2409	return rc;
2410	}
2411
2412
2413	const VDIMAGEBACKEND g_QedBackend =
2414	{
2415	/* u32Version */
2416	VD_IMGBACKEND_VERSION,
2417	/* pszBackendName */
2418	"QED",
2419	/* uBackendCaps */
2420	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2421	/* paFileExtensions */
2422	s_aQedFileExtensions,
2423	/* paConfigInfo */
2424	NULL,
2425	/* pfnProbe */
2426	qedProbe,
2427	/* pfnOpen */
2428	qedOpen,
2429	/* pfnCreate */
2430	qedCreate,
2431	/* pfnRename */
2432	qedRename,
2433	/* pfnClose */
2434	qedClose,
2435	/* pfnRead */
2436	qedRead,
2437	/* pfnWrite */
2438	qedWrite,
2439	/* pfnFlush */
2440	qedFlush,
2441	/* pfnDiscard */
2442	NULL,
2443	/* pfnGetVersion */
2444	qedGetVersion,
2445	/* pfnGetFileSize */
2446	qedGetFileSize,
2447	/* pfnGetPCHSGeometry */
2448	qedGetPCHSGeometry,
2449	/* pfnSetPCHSGeometry */
2450	qedSetPCHSGeometry,
2451	/* pfnGetLCHSGeometry */
2452	qedGetLCHSGeometry,
2453	/* pfnSetLCHSGeometry */
2454	qedSetLCHSGeometry,
2455	/* pfnQueryRegions */
2456	qedQueryRegions,
2457	/* pfnRegionListRelease */
2458	qedRegionListRelease,
2459	/* pfnGetImageFlags */
2460	qedGetImageFlags,
2461	/* pfnGetOpenFlags */
2462	qedGetOpenFlags,
2463	/* pfnSetOpenFlags */
2464	qedSetOpenFlags,
2465	/* pfnGetComment */
2466	qedGetComment,
2467	/* pfnSetComment */
2468	qedSetComment,
2469	/* pfnGetUuid */
2470	qedGetUuid,
2471	/* pfnSetUuid */
2472	qedSetUuid,
2473	/* pfnGetModificationUuid */
2474	qedGetModificationUuid,
2475	/* pfnSetModificationUuid */
2476	qedSetModificationUuid,
2477	/* pfnGetParentUuid */
2478	qedGetParentUuid,
2479	/* pfnSetParentUuid */
2480	qedSetParentUuid,
2481	/* pfnGetParentModificationUuid */
2482	qedGetParentModificationUuid,
2483	/* pfnSetParentModificationUuid */
2484	qedSetParentModificationUuid,
2485	/* pfnDump */
2486	qedDump,
2487	/* pfnGetTimestamp */
2488	NULL,
2489	/* pfnGetParentTimestamp */
2490	NULL,
2491	/* pfnSetParentTimestamp */
2492	NULL,
2493	/* pfnGetParentFilename */
2494	qedGetParentFilename,
2495	/* pfnSetParentFilename */
2496	qedSetParentFilename,
2497	/* pfnComposeLocation */
2498	genericFileComposeLocation,
2499	/* pfnComposeName */
2500	genericFileComposeName,
2501	/* pfnCompact */
2502	NULL,
2503	/* pfnResize */
2504	qedResize,
2505	/* pfnRepair */
2506	NULL,
2507	/* pfnTraverseMetadata */
2508	NULL,
2509	/* u32Version */
2510	VD_IMGBACKEND_VERSION
2511	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 66486

Download in other formats: