QED.cpp@ 100964

Last change on this file since 100964 was 99739, checked in by vboxsync, 19 months ago
*: doxygen corrections (mostly about removing @returns from functions returning void).
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 86.9 KB

Line
1	/* $Id: QED.cpp 99739 2023-05-11 01:01:08Z vboxsync $ */
2	/** @file
3	* QED - QED Disk image.
4	*
5	* The QED backend implements support for the qemu enhanced disk format (short QED)
6	* The specification for the format is available under http://wiki.qemu.org/Features/QED/Specification
7	*
8	* Missing things to implement:
9	* - compaction
10	* - resizing which requires block relocation (very rare case)
11	*/
12
13	/*
14	* Copyright (C) 2011-2023 Oracle and/or its affiliates.
15	*
16	* This file is part of VirtualBox base platform packages, as
17	* available from https://www.virtualbox.org.
18	*
19	* This program is free software; you can redistribute it and/or
20	* modify it under the terms of the GNU General Public License
21	* as published by the Free Software Foundation, in version 3 of the
22	* License.
23	*
24	* This program is distributed in the hope that it will be useful, but
25	* WITHOUT ANY WARRANTY; without even the implied warranty of
26	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
27	* General Public License for more details.
28	*
29	* You should have received a copy of the GNU General Public License
30	* along with this program; if not, see <https://www.gnu.org/licenses>.
31	*
32	* SPDX-License-Identifier: GPL-3.0-only
33	*/
34
35
36	/*********************************************************************************************************************************
37	* Header Files *
38	*********************************************************************************************************************************/
39	#define LOG_GROUP LOG_GROUP_VD_QED
40	#include <VBox/vd-plugin.h>
41	#include <VBox/err.h>
42
43	#include <VBox/log.h>
44	#include <iprt/asm.h>
45	#include <iprt/assert.h>
46	#include <iprt/string.h>
47	#include <iprt/alloc.h>
48	#include <iprt/path.h>
49	#include <iprt/list.h>
50
51	#include "VDBackends.h"
52	#include "VDBackendsInline.h"
53
54
55	/*********************************************************************************************************************************
56	* Structures in a QED image, little endian *
57	*********************************************************************************************************************************/
58
59	#pragma pack(1)
60	typedef struct QedHeader
61	{
62	/** Magic value. */
63	uint32_t u32Magic;
64	/** Cluster size in bytes. */
65	uint32_t u32ClusterSize;
66	/** Size of L1 and L2 tables in clusters. */
67	uint32_t u32TableSize;
68	/** size of this header structure in clusters. */
69	uint32_t u32HeaderSize;
70	/** Features used for the image. */
71	uint64_t u64FeatureFlags;
72	/** Compatibility features used for the image. */
73	uint64_t u64CompatFeatureFlags;
74	/** Self resetting feature bits. */
75	uint64_t u64AutoresetFeatureFlags;
76	/** Offset of the L1 table in bytes. */
77	uint64_t u64OffL1Table;
78	/** Logical image size as seen by the guest. */
79	uint64_t u64Size;
80	/** Offset of the backing filename in bytes. */
81	uint32_t u32OffBackingFilename;
82	/** Size of the backing filename. */
83	uint32_t u32BackingFilenameSize;
84	} QedHeader;
85	#pragma pack()
86	/** Pointer to a on disk QED header. */
87	typedef QedHeader *PQedHeader;
88
89	/** QED magic value. */
90	#define QED_MAGIC UINT32_C(0x00444551) /* QED\0 */
91	/** Cluster size minimum. */
92	#define QED_CLUSTER_SIZE_MIN RT_BIT(12)
93	/** Cluster size maximum. */
94	#define QED_CLUSTER_SIZE_MAX RT_BIT(26)
95	/** L1 and L2 Table size minimum. */
96	#define QED_TABLE_SIZE_MIN 1
97	/** L1 and L2 Table size maximum. */
98	#define QED_TABLE_SIZE_MAX 16
99
100	/** QED default cluster size when creating an image. */
101	#define QED_CLUSTER_SIZE_DEFAULT (64 * _1K)
102	/** The default table size in clusters. */
103	#define QED_TABLE_SIZE_DEFAULT 4
104
105	/** Feature flags.
106	* @{
107	*/
108	/** Image uses a backing file to provide data for unallocated clusters. */
109	#define QED_FEATURE_BACKING_FILE RT_BIT_64(0)
110	/** Image needs checking before use. */
111	#define QED_FEATURE_NEED_CHECK RT_BIT_64(1)
112	/** Don't probe for format of the backing file, treat as raw image. */
113	#define QED_FEATURE_BACKING_FILE_NO_PROBE RT_BIT_64(2)
114	/** Mask of valid features. */
115	#define QED_FEATURE_MASK (QED_FEATURE_BACKING_FILE \| QED_FEATURE_NEED_CHECK \| QED_FEATURE_BACKING_FILE_NO_PROBE)
116	/** @} */
117
118	/** Compatibility feature flags.
119	* @{
120	*/
121	/** Mask of valid compatibility features. */
122	#define QED_COMPAT_FEATURE_MASK (0)
123	/** @} */
124
125	/** Autoreset feature flags.
126	* @{
127	*/
128	/** Mask of valid autoreset features. */
129	#define QED_AUTORESET_FEATURE_MASK (0)
130	/** @} */
131
132
133	/*********************************************************************************************************************************
134	* Constants And Macros, Structures and Typedefs *
135	*********************************************************************************************************************************/
136
137	/**
138	* QED L2 cache entry.
139	*/
140	typedef struct QEDL2CACHEENTRY
141	{
142	/** List node for the search list. */
143	RTLISTNODE NodeSearch;
144	/** List node for the LRU list. */
145	RTLISTNODE NodeLru;
146	/** Reference counter. */
147	uint32_t cRefs;
148	/** The offset of the L2 table, used as search key. */
149	uint64_t offL2Tbl;
150	/** Pointer to the cached L2 table. */
151	uint64_t *paL2Tbl;
152	} QEDL2CACHEENTRY, *PQEDL2CACHEENTRY;
153
154	/** Maximum amount of memory the cache is allowed to use. */
155	#define QED_L2_CACHE_MEMORY_MAX (2*_1M)
156
157	/**
158	* QED image data structure.
159	*/
160	typedef struct QEDIMAGE
161	{
162	/** Image name. */
163	const char *pszFilename;
164	/** Storage handle. */
165	PVDIOSTORAGE pStorage;
166
167	/** Pointer to the per-disk VD interface list. */
168	PVDINTERFACE pVDIfsDisk;
169	/** Pointer to the per-image VD interface list. */
170	PVDINTERFACE pVDIfsImage;
171	/** Error interface. */
172	PVDINTERFACEERROR pIfError;
173	/** I/O interface. */
174	PVDINTERFACEIOINT pIfIo;
175
176	/** Open flags passed by VBoxHD layer. */
177	unsigned uOpenFlags;
178	/** Image flags defined during creation or determined during open. */
179	unsigned uImageFlags;
180	/** Total size of the image. */
181	uint64_t cbSize;
182	/** Physical geometry of this image. */
183	VDGEOMETRY PCHSGeometry;
184	/** Logical geometry of this image. */
185	VDGEOMETRY LCHSGeometry;
186
187	/** Filename of the backing file if any. */
188	char *pszBackingFilename;
189	/** Offset of the filename in the image. */
190	uint32_t offBackingFilename;
191	/** Size of the backing filename excluding \0. */
192	uint32_t cbBackingFilename;
193
194	/** Size of the image, multiple of clusters. */
195	uint64_t cbImage;
196	/** Cluster size in bytes. */
197	uint32_t cbCluster;
198	/** Number of entries in the L1 and L2 table. */
199	uint32_t cTableEntries;
200	/** Size of an L1 or L2 table rounded to the next cluster size. */
201	uint32_t cbTable;
202	/** Pointer to the L1 table. */
203	uint64_t *paL1Table;
204	/** Offset of the L1 table. */
205	uint64_t offL1Table;
206
207	/** Offset mask for a cluster. */
208	uint64_t fOffsetMask;
209	/** L1 table mask to get the L1 index. */
210	uint64_t fL1Mask;
211	/** Number of bits to shift to get the L1 index. */
212	uint32_t cL1Shift;
213	/** L2 table mask to get the L2 index. */
214	uint64_t fL2Mask;
215	/** Number of bits to shift to get the L2 index. */
216	uint32_t cL2Shift;
217
218	/** Pointer to the L2 table we are currently allocating
219	* (can be only one at a time). */
220	PQEDL2CACHEENTRY pL2TblAlloc;
221
222	/** Memory occupied by the L2 table cache. */
223	size_t cbL2Cache;
224	/** The sorted L2 entry list used for searching. */
225	RTLISTNODE ListSearch;
226	/** The LRU L2 entry list used for eviction. */
227	RTLISTNODE ListLru;
228	/** The static region list. */
229	VDREGIONLIST RegionList;
230	} QEDIMAGE, *PQEDIMAGE;
231
232	/**
233	* State of the async cluster allocation.
234	*/
235	typedef enum QEDCLUSTERASYNCALLOCSTATE
236	{
237	/** Invalid. */
238	QEDCLUSTERASYNCALLOCSTATE_INVALID = 0,
239	/** L2 table allocation. */
240	QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC,
241	/** Link L2 table into L1. */
242	QEDCLUSTERASYNCALLOCSTATE_L2_LINK,
243	/** Allocate user data cluster. */
244	QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC,
245	/** Link user data cluster. */
246	QEDCLUSTERASYNCALLOCSTATE_USER_LINK,
247	/** 32bit blowup. */
248	QEDCLUSTERASYNCALLOCSTATE_32BIT_HACK = 0x7fffffff
249	} QEDCLUSTERASYNCALLOCSTATE, *PQEDCLUSTERASYNCALLOCSTATE;
250
251	/**
252	* Data needed to track async cluster allocation.
253	*/
254	typedef struct QEDCLUSTERASYNCALLOC
255	{
256	/** The state of the cluster allocation. */
257	QEDCLUSTERASYNCALLOCSTATE enmAllocState;
258	/** Old image size to rollback in case of an error. */
259	uint64_t cbImageOld;
260	/** L1 index to link if any. */
261	uint32_t idxL1;
262	/** L2 index to link, required in any case. */
263	uint32_t idxL2;
264	/** Start offset of the allocated cluster. */
265	uint64_t offClusterNew;
266	/** L2 cache entry if a L2 table is allocated. */
267	PQEDL2CACHEENTRY pL2Entry;
268	/** Number of bytes to write. */
269	size_t cbToWrite;
270	} QEDCLUSTERASYNCALLOC, *PQEDCLUSTERASYNCALLOC;
271
272
273	/*********************************************************************************************************************************
274	* Static Variables *
275	*********************************************************************************************************************************/
276
277	/** NULL-terminated array of supported file extensions. */
278	static const VDFILEEXTENSION s_aQedFileExtensions[] =
279	{
280	{"qed", VDTYPE_HDD},
281	{NULL, VDTYPE_INVALID}
282	};
283
284
285	/*********************************************************************************************************************************
286	* Internal Functions *
287	*********************************************************************************************************************************/
288
289	/**
290	* Converts the image header to the host endianess and performs basic checks.
291	*
292	* @returns Whether the given header is valid or not.
293	* @param pHeader Pointer to the header to convert.
294	*/
295	static bool qedHdrConvertToHostEndianess(PQedHeader pHeader)
296	{
297	pHeader->u32Magic = RT_LE2H_U32(pHeader->u32Magic);
298	pHeader->u32ClusterSize = RT_LE2H_U32(pHeader->u32ClusterSize);
299	pHeader->u32TableSize = RT_LE2H_U32(pHeader->u32TableSize);
300	pHeader->u32HeaderSize = RT_LE2H_U32(pHeader->u32HeaderSize);
301	pHeader->u64FeatureFlags = RT_LE2H_U64(pHeader->u64FeatureFlags);
302	pHeader->u64CompatFeatureFlags = RT_LE2H_U64(pHeader->u64CompatFeatureFlags);
303	pHeader->u64AutoresetFeatureFlags = RT_LE2H_U64(pHeader->u64AutoresetFeatureFlags);
304	pHeader->u64OffL1Table = RT_LE2H_U64(pHeader->u64OffL1Table);
305	pHeader->u64Size = RT_LE2H_U64(pHeader->u64Size);
306	pHeader->u32OffBackingFilename = RT_LE2H_U32(pHeader->u32OffBackingFilename);
307	pHeader->u32BackingFilenameSize = RT_LE2H_U32(pHeader->u32BackingFilenameSize);
308
309	if (RT_UNLIKELY(pHeader->u32Magic != QED_MAGIC))
310	return false;
311	if (RT_UNLIKELY( pHeader->u32ClusterSize < QED_CLUSTER_SIZE_MIN
312	\|\| pHeader->u32ClusterSize > QED_CLUSTER_SIZE_MAX))
313	return false;
314	if (RT_UNLIKELY( pHeader->u32TableSize < QED_TABLE_SIZE_MIN
315	\|\| pHeader->u32TableSize > QED_TABLE_SIZE_MAX))
316	return false;
317	if (RT_UNLIKELY(pHeader->u64Size % 512 != 0))
318	return false;
319	if (RT_UNLIKELY( pHeader->u64FeatureFlags & QED_FEATURE_BACKING_FILE
320	&& ( pHeader->u32BackingFilenameSize == 0
321	\|\| pHeader->u32BackingFilenameSize == UINT32_MAX)))
322	return false;
323
324	return true;
325	}
326
327	/**
328	* Creates a QED header from the given image state.
329	*
330	* @param pImage Image instance data.
331	* @param pHeader Pointer to the header to convert.
332	*/
333	static void qedHdrConvertFromHostEndianess(PQEDIMAGE pImage, PQedHeader pHeader)
334	{
335	pHeader->u32Magic = RT_H2LE_U32(QED_MAGIC);
336	pHeader->u32ClusterSize = RT_H2LE_U32(pImage->cbCluster);
337	pHeader->u32TableSize = RT_H2LE_U32(pImage->cbTable / pImage->cbCluster);
338	pHeader->u32HeaderSize = RT_H2LE_U32(1);
339	pHeader->u64FeatureFlags = RT_H2LE_U64(pImage->pszBackingFilename ? QED_FEATURE_BACKING_FILE : UINT64_C(0));
340	pHeader->u64CompatFeatureFlags = RT_H2LE_U64(UINT64_C(0));
341	pHeader->u64AutoresetFeatureFlags = RT_H2LE_U64(UINT64_C(0));
342	pHeader->u64OffL1Table = RT_H2LE_U64(pImage->offL1Table);
343	pHeader->u64Size = RT_H2LE_U64(pImage->cbSize);
344	pHeader->u32OffBackingFilename = RT_H2LE_U32(pImage->offBackingFilename);
345	pHeader->u32BackingFilenameSize = RT_H2LE_U32(pImage->cbBackingFilename);
346	}
347
348	/**
349	* Convert table entries from little endian to host endianess.
350	*
351	* @param paTbl Pointer to the table.
352	* @param cEntries Number of entries in the table.
353	*/
354	static void qedTableConvertToHostEndianess(uint64_t *paTbl, uint32_t cEntries)
355	{
356	while (cEntries-- > 0)
357	{
358	paTbl = RT_LE2H_U64(paTbl);
359	paTbl++;
360	}
361	}
362
363	#if defined(RT_BIG_ENDIAN)
364	/**
365	* Convert table entries from host to little endian format.
366	*
367	* @param paTblImg Pointer to the table which will store the little endian table.
368	* @param paTbl The source table to convert.
369	* @param cEntries Number of entries in the table.
370	*/
371	static void qedTableConvertFromHostEndianess(uint64_t paTblImg, uint64_t const paTbl,
372	uint32_t cEntries)
373	{
374	while (cEntries-- > 0)
375	{
376	paTblImg = RT_H2LE_U64(paTbl);
377	paTbl++;
378	paTblImg++;
379	}
380	}
381	#endif
382
383	/**
384	* Creates the L2 table cache.
385	*
386	* @returns VBox status code.
387	* @param pImage The image instance data.
388	*/
389	static int qedL2TblCacheCreate(PQEDIMAGE pImage)
390	{
391	pImage->cbL2Cache = 0;
392	RTListInit(&pImage->ListSearch);
393	RTListInit(&pImage->ListLru);
394
395	return VINF_SUCCESS;
396	}
397
398	/**
399	* Destroys the L2 table cache.
400	*
401	* @param pImage The image instance data.
402	*/
403	static void qedL2TblCacheDestroy(PQEDIMAGE pImage)
404	{
405	PQEDL2CACHEENTRY pL2Entry;
406	PQEDL2CACHEENTRY pL2Next;
407	RTListForEachSafe(&pImage->ListSearch, pL2Entry, pL2Next, QEDL2CACHEENTRY, NodeSearch)
408	{
409	Assert(!pL2Entry->cRefs);
410
411	RTListNodeRemove(&pL2Entry->NodeSearch);
412	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
413	RTMemFree(pL2Entry);
414	}
415
416	pImage->cbL2Cache = 0;
417	RTListInit(&pImage->ListSearch);
418	RTListInit(&pImage->ListLru);
419	}
420
421	/**
422	* Returns the L2 table matching the given offset or NULL if none could be found.
423	*
424	* @returns Pointer to the L2 table cache entry or NULL.
425	* @param pImage The image instance data.
426	* @param offL2Tbl Offset of the L2 table to search for.
427	*/
428	static PQEDL2CACHEENTRY qedL2TblCacheRetain(PQEDIMAGE pImage, uint64_t offL2Tbl)
429	{
430	if ( pImage->pL2TblAlloc
431	&& pImage->pL2TblAlloc->offL2Tbl == offL2Tbl)
432	{
433	pImage->pL2TblAlloc->cRefs++;
434	return pImage->pL2TblAlloc;
435	}
436
437	PQEDL2CACHEENTRY pL2Entry;
438	RTListForEach(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch)
439	{
440	if (pL2Entry->offL2Tbl == offL2Tbl)
441	break;
442	}
443
444	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
445	{
446	/* Update LRU list. */
447	RTListNodeRemove(&pL2Entry->NodeLru);
448	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
449	pL2Entry->cRefs++;
450	return pL2Entry;
451	}
452	else
453	return NULL;
454	}
455
456	/**
457	* Releases a L2 table cache entry.
458	*
459	* @param pL2Entry The L2 cache entry.
460	*/
461	static void qedL2TblCacheEntryRelease(PQEDL2CACHEENTRY pL2Entry)
462	{
463	Assert(pL2Entry->cRefs > 0);
464	pL2Entry->cRefs--;
465	}
466
467	/**
468	* Allocates a new L2 table from the cache evicting old entries if required.
469	*
470	* @returns Pointer to the L2 cache entry or NULL.
471	* @param pImage The image instance data.
472	*/
473	static PQEDL2CACHEENTRY qedL2TblCacheEntryAlloc(PQEDIMAGE pImage)
474	{
475	PQEDL2CACHEENTRY pL2Entry = NULL;
476
477	if (pImage->cbL2Cache + pImage->cbTable <= QED_L2_CACHE_MEMORY_MAX)
478	{
479	/* Add a new entry. */
480	pL2Entry = (PQEDL2CACHEENTRY)RTMemAllocZ(sizeof(QEDL2CACHEENTRY));
481	if (pL2Entry)
482	{
483	pL2Entry->paL2Tbl = (uint64_t *)RTMemPageAllocZ(pImage->cbTable);
484	if (RT_UNLIKELY(!pL2Entry->paL2Tbl))
485	{
486	RTMemFree(pL2Entry);
487	pL2Entry = NULL;
488	}
489	else
490	{
491	pL2Entry->cRefs = 1;
492	pImage->cbL2Cache += pImage->cbTable;
493	}
494	}
495	}
496	else
497	{
498	/* Evict the last not in use entry and use it */
499	Assert(!RTListIsEmpty(&pImage->ListLru));
500
501	RTListForEachReverse(&pImage->ListLru, pL2Entry, QEDL2CACHEENTRY, NodeLru)
502	{
503	if (!pL2Entry->cRefs)
504	break;
505	}
506
507	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
508	{
509	RTListNodeRemove(&pL2Entry->NodeSearch);
510	RTListNodeRemove(&pL2Entry->NodeLru);
511	pL2Entry->offL2Tbl = 0;
512	pL2Entry->cRefs = 1;
513	}
514	else
515	pL2Entry = NULL;
516	}
517
518	return pL2Entry;
519	}
520
521	/**
522	* Frees a L2 table cache entry.
523	*
524	* @param pImage The image instance data.
525	* @param pL2Entry The L2 cache entry to free.
526	*/
527	static void qedL2TblCacheEntryFree(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
528	{
529	Assert(!pL2Entry->cRefs);
530	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
531	RTMemFree(pL2Entry);
532
533	pImage->cbL2Cache -= pImage->cbTable;
534	}
535
536	/**
537	* Inserts an entry in the L2 table cache.
538	*
539	* @param pImage The image instance data.
540	* @param pL2Entry The L2 cache entry to insert.
541	*/
542	static void qedL2TblCacheEntryInsert(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
543	{
544	Assert(pL2Entry->offL2Tbl > 0);
545
546	/* Insert at the top of the LRU list. */
547	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
548
549	if (RTListIsEmpty(&pImage->ListSearch))
550	{
551	RTListAppend(&pImage->ListSearch, &pL2Entry->NodeSearch);
552	}
553	else
554	{
555	/* Insert into search list. */
556	PQEDL2CACHEENTRY pIt;
557	pIt = RTListGetFirst(&pImage->ListSearch, QEDL2CACHEENTRY, NodeSearch);
558	if (pIt->offL2Tbl > pL2Entry->offL2Tbl)
559	RTListPrepend(&pImage->ListSearch, &pL2Entry->NodeSearch);
560	else
561	{
562	bool fInserted = false;
563
564	RTListForEach(&pImage->ListSearch, pIt, QEDL2CACHEENTRY, NodeSearch)
565	{
566	Assert(pIt->offL2Tbl != pL2Entry->offL2Tbl);
567	if (pIt->offL2Tbl < pL2Entry->offL2Tbl)
568	{
569	RTListNodeInsertAfter(&pIt->NodeSearch, &pL2Entry->NodeSearch);
570	fInserted = true;
571	break;
572	}
573	}
574	Assert(fInserted);
575	}
576	}
577	}
578
579	/**
580	* Fetches the L2 from the given offset trying the LRU cache first and
581	* reading it from the image after a cache miss - version for async I/O.
582	*
583	* @returns VBox status code.
584	* @param pImage Image instance data.
585	* @param pIoCtx The I/O context.
586	* @param offL2Tbl The offset of the L2 table in the image.
587	* @param ppL2Entry Where to store the L2 table on success.
588	*/
589	static int qedL2TblCacheFetchAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
590	uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
591	{
592	int rc = VINF_SUCCESS;
593
594	/* Try to fetch the L2 table from the cache first. */
595	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
596	if (!pL2Entry)
597	{
598	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
599
600	if (pL2Entry)
601	{
602	/* Read from the image. */
603	PVDMETAXFER pMetaXfer;
604
605	pL2Entry->offL2Tbl = offL2Tbl;
606	rc = vdIfIoIntFileReadMeta(pImage->pIfIo, pImage->pStorage,
607	offL2Tbl, pL2Entry->paL2Tbl,
608	pImage->cbTable, pIoCtx,
609	&pMetaXfer, NULL, NULL);
610	if (RT_SUCCESS(rc))
611	{
612	vdIfIoIntMetaXferRelease(pImage->pIfIo, pMetaXfer);
613	#if defined(RT_BIG_ENDIAN)
614	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
615	#endif
616	qedL2TblCacheEntryInsert(pImage, pL2Entry);
617	}
618	else
619	{
620	qedL2TblCacheEntryRelease(pL2Entry);
621	qedL2TblCacheEntryFree(pImage, pL2Entry);
622	}
623	}
624	else
625	rc = VERR_NO_MEMORY;
626	}
627
628	if (RT_SUCCESS(rc))
629	*ppL2Entry = pL2Entry;
630
631	return rc;
632	}
633
634	/**
635	* Return power of 2 or 0 if num error.
636	*
637	* @returns The power of 2 or 0 if the given number is not a power of 2.
638	* @param u32 The number.
639	*/
640	static uint32_t qedGetPowerOfTwo(uint32_t u32)
641	{
642	if (u32 == 0)
643	return 0;
644	uint32_t uPower2 = 0;
645	while ((u32 & 1) == 0)
646	{
647	u32 >>= 1;
648	uPower2++;
649	}
650	return u32 == 1 ? uPower2 : 0;
651	}
652
653	/**
654	* Sets the L1, L2 and offset bitmasks and L1 and L2 bit shift members.
655	*
656	* @param pImage The image instance data.
657	*/
658	static void qedTableMasksInit(PQEDIMAGE pImage)
659	{
660	uint32_t cClusterBits, cTableBits;
661
662	cClusterBits = qedGetPowerOfTwo(pImage->cbCluster);
663	cTableBits = qedGetPowerOfTwo(pImage->cTableEntries);
664
665	Assert(cClusterBits + 2 * cTableBits <= 64);
666
667	pImage->fOffsetMask = ((uint64_t)pImage->cbCluster - 1);
668	pImage->fL2Mask = ((uint64_t)pImage->cTableEntries - 1) << cClusterBits;
669	pImage->cL2Shift = cClusterBits;
670	pImage->fL1Mask = ((uint64_t)pImage->cTableEntries - 1) << (cClusterBits + cTableBits);
671	pImage->cL1Shift = cClusterBits + cTableBits;
672	}
673
674	/**
675	* Converts a given logical offset into the
676	*
677	* @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	RTStrFree(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 = RTStrAlloc(Header.u32BackingFilenameSize + 1); /* +1 for \0 terminator. */
1177	if (pImage->pszBackingFilename)
1178	{
1179	RT_BZERO(pImage->pszBackingFilename, Header.u32BackingFilenameSize + 1);
1180	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1181	pImage->offBackingFilename = Header.u32OffBackingFilename;
1182	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1183	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1184	Header.u32BackingFilenameSize);
1185	if (RT_SUCCESS(rc))
1186	rc = RTStrValidateEncoding(pImage->pszBackingFilename);
1187	}
1188	else
1189	rc = VERR_NO_STR_MEMORY;
1190	}
1191
1192	if (RT_SUCCESS(rc))
1193	{
1194	pImage->cbImage = cbFile;
1195	pImage->cbCluster = Header.u32ClusterSize;
1196	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1197	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1198	pImage->offL1Table = Header.u64OffL1Table;
1199	pImage->cbSize = Header.u64Size;
1200	qedTableMasksInit(pImage);
1201
1202	/* Allocate L1 table. */
1203	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1204	if (pImage->paL1Table)
1205	{
1206	/* Read from the image. */
1207	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1208	pImage->offL1Table, pImage->paL1Table,
1209	pImage->cbTable);
1210	if (RT_SUCCESS(rc))
1211	{
1212	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1213
1214	/* If the consistency check succeeded, clear the flag by flushing the image. */
1215	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1216	rc = qedFlushImage(pImage);
1217	}
1218	else
1219	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1220	N_("Qed: Reading the L1 table for image '%s' failed"),
1221	pImage->pszFilename);
1222	}
1223	else
1224	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1225	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1226	pImage->pszFilename);
1227	}
1228	}
1229	else
1230	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1231	N_("Qed: The image '%s' makes use of unsupported features"),
1232	pImage->pszFilename);
1233	}
1234	else if (RT_SUCCESS(rc))
1235	rc = VERR_VD_GEN_INVALID_HEADER;
1236	}
1237	else if (RT_SUCCESS(rc))
1238	rc = VERR_VD_GEN_INVALID_HEADER;
1239	}
1240	/* else: Do NOT signal an appropriate error here, as the VD layer has the
1241	* choice of retrying the open if it failed. */
1242	}
1243	else
1244	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1245	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1246	pImage->pszFilename);
1247
1248	if (RT_SUCCESS(rc))
1249	{
1250	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1251	pImage->RegionList.fFlags = 0;
1252	pImage->RegionList.cRegions = 1;
1253
1254	pRegion->offRegion = 0; /* Disk start. */
1255	pRegion->cbBlock = 512;
1256	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1257	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1258	pRegion->cbData = 512;
1259	pRegion->cbMetadata = 0;
1260	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1261	}
1262	else
1263	qedFreeImage(pImage, false);
1264	return rc;
1265	}
1266
1267	/**
1268	* Internal: Create a qed image.
1269	*/
1270	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1271	unsigned uImageFlags, const char *pszComment,
1272	PCVDGEOMETRY pPCHSGeometry,
1273	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1274	PVDINTERFACEPROGRESS pIfProgress,
1275	unsigned uPercentStart, unsigned uPercentSpan)
1276	{
1277	RT_NOREF1(pszComment);
1278	int rc;
1279
1280	if (!(uImageFlags & VD_IMAGE_FLAGS_FIXED))
1281	{
1282	rc = qedL2TblCacheCreate(pImage);
1283	if (RT_SUCCESS(rc))
1284	{
1285	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1286	pImage->uImageFlags = uImageFlags;
1287	pImage->PCHSGeometry = *pPCHSGeometry;
1288	pImage->LCHSGeometry = *pLCHSGeometry;
1289
1290	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1291	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1292	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1293
1294	/* Create image file. */
1295	uint32_t fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1296	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1297	if (RT_SUCCESS(rc))
1298	{
1299	/* Init image state. */
1300	pImage->cbSize = cbSize;
1301	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1302	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1303	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1304	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1305	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1306	pImage->cbBackingFilename = 0;
1307	pImage->offBackingFilename = 0;
1308	qedTableMasksInit(pImage);
1309
1310	/* Init L1 table. */
1311	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1312	if (RT_LIKELY(pImage->paL1Table))
1313	{
1314	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan * 98 / 100);
1315	rc = qedFlushImage(pImage);
1316	}
1317	else
1318	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1319	pImage->pszFilename);
1320	}
1321	else
1322	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1323	}
1324	else
1325	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1326	pImage->pszFilename);
1327	}
1328	else
1329	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1330
1331	if (RT_SUCCESS(rc))
1332	{
1333	PVDREGIONDESC pRegion = &pImage->RegionList.aRegions[0];
1334	pImage->RegionList.fFlags = 0;
1335	pImage->RegionList.cRegions = 1;
1336
1337	pRegion->offRegion = 0; /* Disk start. */
1338	pRegion->cbBlock = 512;
1339	pRegion->enmDataForm = VDREGIONDATAFORM_RAW;
1340	pRegion->enmMetadataForm = VDREGIONMETADATAFORM_NONE;
1341	pRegion->cbData = 512;
1342	pRegion->cbMetadata = 0;
1343	pRegion->cRegionBlocksOrBytes = pImage->cbSize;
1344
1345	vdIfProgress(pIfProgress, uPercentStart + uPercentSpan);
1346	}
1347	else
1348	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1349
1350	return rc;
1351	}
1352
1353	/**
1354	* Rollback anything done during async cluster allocation.
1355	*
1356	* @returns VBox status code.
1357	* @param pImage The image instance data.
1358	* @param pIoCtx The I/O context.
1359	* @param pClusterAlloc The cluster allocation to rollback.
1360	*/
1361	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1362	{
1363	RT_NOREF1(pIoCtx);
1364	int rc = VINF_SUCCESS;
1365
1366	switch (pClusterAlloc->enmAllocState)
1367	{
1368	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1369	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1370	{
1371	/* Revert the L1 table entry */
1372	pImage->paL1Table[pClusterAlloc->idxL1] = 0;
1373	pImage->pL2TblAlloc = NULL;
1374
1375	/* Assumption right now is that the L1 table is not modified on storage if the link fails. */
1376	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1377	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1378	Assert(!pClusterAlloc->pL2Entry->cRefs);
1379	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1380	break;
1381	}
1382	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1383	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1384	{
1385	/* Assumption right now is that the L2 table is not modified if the link fails. */
1386	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = 0;
1387	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1388	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1389	break;
1390	}
1391	default:
1392	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1393	rc = VERR_INVALID_STATE;
1394	}
1395
1396	RTMemFree(pClusterAlloc);
1397	return rc;
1398	}
1399
1400	/**
1401	* Updates the state of the async cluster allocation.
1402	*
1403	* @returns VBox status code.
1404	* @param pBackendData The opaque backend data.
1405	* @param pIoCtx I/O context associated with this request.
1406	* @param pvUser Opaque user data passed during a read/write request.
1407	* @param rcReq Status code for the completed request.
1408	*/
1409	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1410	{
1411	int rc = VINF_SUCCESS;
1412	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1413	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1414
1415	if (RT_FAILURE(rcReq))
1416	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1417
1418	AssertPtr(pClusterAlloc->pL2Entry);
1419
1420	switch (pClusterAlloc->enmAllocState)
1421	{
1422	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1423	{
1424	/* Update the link in the in memory L1 table now. */
1425	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1426
1427	/* Update the link in the on disk L1 table now. */
1428	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1429	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1430	qedAsyncClusterAllocUpdate, pClusterAlloc);
1431	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1432	break;
1433	else if (RT_FAILURE(rc))
1434	{
1435	/* Rollback. */
1436	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1437	break;
1438	}
1439	}
1440	RT_FALL_THRU();
1441	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1442	{
1443	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1444	uint64_t offData = qedClusterAllocate(pImage, 1);
1445
1446	pImage->pL2TblAlloc = NULL;
1447	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1448
1449	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1450	pClusterAlloc->cbImageOld = offData;
1451	pClusterAlloc->offClusterNew = offData;
1452
1453	/* Write data. */
1454	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1455	offData, pIoCtx, pClusterAlloc->cbToWrite,
1456	qedAsyncClusterAllocUpdate, pClusterAlloc);
1457	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1458	break;
1459	else if (RT_FAILURE(rc))
1460	{
1461	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1462	RTMemFree(pClusterAlloc);
1463	break;
1464	}
1465	}
1466	RT_FALL_THRU();
1467	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1468	{
1469	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1470	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1471
1472	/* Link L2 table and update it. */
1473	rc = qedTblWrite(pImage, pIoCtx, pImage->paL1Table[pClusterAlloc->idxL1],
1474	pClusterAlloc->pL2Entry->paL2Tbl,
1475	qedAsyncClusterAllocUpdate, pClusterAlloc);
1476	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1477	break;
1478	else if (RT_FAILURE(rc))
1479	{
1480	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1481	RTMemFree(pClusterAlloc);
1482	break;
1483	}
1484	}
1485	RT_FALL_THRU();
1486	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1487	{
1488	/* Everything done without errors, signal completion. */
1489	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1490	RTMemFree(pClusterAlloc);
1491	rc = VINF_SUCCESS;
1492	break;
1493	}
1494	default:
1495	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1496	pClusterAlloc->enmAllocState));
1497	}
1498
1499	return rc;
1500	}
1501
1502	/** @copydoc VDIMAGEBACKEND::pfnProbe */
1503	static DECLCALLBACK(int) qedProbe(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1504	PVDINTERFACE pVDIfsImage, VDTYPE enmDesiredType, VDTYPE *penmType)
1505	{
1506	RT_NOREF(pVDIfsDisk, enmDesiredType);
1507	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1508	PVDIOSTORAGE pStorage = NULL;
1509	int rc = VINF_SUCCESS;
1510
1511	/* Get I/O interface. */
1512	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1513	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1514	AssertPtrReturn(pszFilename, VERR_INVALID_POINTER);
1515	AssertReturn(*pszFilename != '\0', VERR_INVALID_PARAMETER);
1516
1517
1518	/*
1519	* Open the file and read the footer.
1520	*/
1521	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1522	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1523	false /* fCreate */),
1524	&pStorage);
1525	if (RT_SUCCESS(rc))
1526	{
1527	uint64_t cbFile;
1528
1529	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1530	if ( RT_SUCCESS(rc)
1531	&& cbFile > sizeof(QedHeader))
1532	{
1533	QedHeader Header;
1534
1535	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1536	if ( RT_SUCCESS(rc)
1537	&& qedHdrConvertToHostEndianess(&Header))
1538	*penmType = VDTYPE_HDD;
1539	else
1540	rc = VERR_VD_GEN_INVALID_HEADER;
1541	}
1542	else
1543	rc = VERR_VD_GEN_INVALID_HEADER;
1544	}
1545
1546	if (pStorage)
1547	vdIfIoIntFileClose(pIfIo, pStorage);
1548
1549	LogFlowFunc(("returns %Rrc\n", rc));
1550	return rc;
1551	}
1552
1553	/** @copydoc VDIMAGEBACKEND::pfnOpen */
1554	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1555	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1556	VDTYPE enmType, void **ppBackendData)
1557	{
1558	RT_NOREF1(enmType); /*< @todo r=klaus make use of the type info. /
1559
1560	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n",
1561	pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1562	int rc;
1563
1564	/* Check open flags. All valid flags are supported. */
1565	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1566	AssertPtrReturn(pszFilename, VERR_INVALID_POINTER);
1567	AssertReturn(*pszFilename != '\0', VERR_INVALID_PARAMETER);
1568
1569
1570	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1571	if (RT_LIKELY(pImage))
1572	{
1573	pImage->pszFilename = pszFilename;
1574	pImage->pStorage = NULL;
1575	pImage->pVDIfsDisk = pVDIfsDisk;
1576	pImage->pVDIfsImage = pVDIfsImage;
1577
1578	rc = qedOpenImage(pImage, uOpenFlags);
1579	if (RT_SUCCESS(rc))
1580	*ppBackendData = pImage;
1581	else
1582	RTMemFree(pImage);
1583	}
1584	else
1585	rc = VERR_NO_MEMORY;
1586
1587	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1588	return rc;
1589	}
1590
1591	/** @copydoc VDIMAGEBACKEND::pfnCreate */
1592	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1593	unsigned uImageFlags, const char *pszComment,
1594	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1595	PCRTUUID pUuid, unsigned uOpenFlags,
1596	unsigned uPercentStart, unsigned uPercentSpan,
1597	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1598	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1599	void **ppBackendData)
1600	{
1601	RT_NOREF1(pUuid);
1602	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",
1603	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1604	int rc;
1605
1606	/* Check the VD container type. */
1607	if (enmType != VDTYPE_HDD)
1608	return VERR_VD_INVALID_TYPE;
1609
1610	/* Check open flags. All valid flags are supported. */
1611	AssertReturn(!(uOpenFlags & ~VD_OPEN_FLAGS_MASK), VERR_INVALID_PARAMETER);
1612	AssertPtrReturn(pszFilename, VERR_INVALID_POINTER);
1613	AssertReturn(*pszFilename != '\0', VERR_INVALID_PARAMETER);
1614	AssertPtrReturn(pPCHSGeometry, VERR_INVALID_POINTER);
1615	AssertPtrReturn(pLCHSGeometry, VERR_INVALID_POINTER);
1616
1617	PQEDIMAGE pImage = (PQEDIMAGE)RTMemAllocZ(RT_UOFFSETOF(QEDIMAGE, RegionList.aRegions[1]));
1618	if (RT_LIKELY(pImage))
1619	{
1620	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1621
1622	pImage->pszFilename = pszFilename;
1623	pImage->pStorage = NULL;
1624	pImage->pVDIfsDisk = pVDIfsDisk;
1625	pImage->pVDIfsImage = pVDIfsImage;
1626
1627	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1628	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1629	pIfProgress, uPercentStart, uPercentSpan);
1630	if (RT_SUCCESS(rc))
1631	{
1632	/* So far the image is opened in read/write mode. Make sure the
1633	* image is opened in read-only mode if the caller requested that. */
1634	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1635	{
1636	qedFreeImage(pImage, false);
1637	rc = qedOpenImage(pImage, uOpenFlags);
1638	}
1639
1640	if (RT_SUCCESS(rc))
1641	*ppBackendData = pImage;
1642	}
1643
1644	if (RT_FAILURE(rc))
1645	RTMemFree(pImage);
1646	}
1647	else
1648	rc = VERR_NO_MEMORY;
1649
1650	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1651	return rc;
1652	}
1653
1654	/** @copydoc VDIMAGEBACKEND::pfnRename */
1655	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1656	{
1657	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1658	int rc = VINF_SUCCESS;
1659	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1660
1661	/* Check arguments. */
1662	AssertReturn((pImage && pszFilename && *pszFilename), VERR_INVALID_PARAMETER);
1663
1664	/* Close the image. */
1665	rc = qedFreeImage(pImage, false);
1666	if (RT_SUCCESS(rc))
1667	{
1668	/* Rename the file. */
1669	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1670	if (RT_SUCCESS(rc))
1671	{
1672	/* Update pImage with the new information. */
1673	pImage->pszFilename = pszFilename;
1674
1675	/* Open the old image with new name. */
1676	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1677	}
1678	else
1679	{
1680	/* The move failed, try to reopen the original image. */
1681	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1682	if (RT_FAILURE(rc2))
1683	rc = rc2;
1684	}
1685	}
1686
1687	LogFlowFunc(("returns %Rrc\n", rc));
1688	return rc;
1689	}
1690
1691	/** @copydoc VDIMAGEBACKEND::pfnClose */
1692	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1693	{
1694	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1695	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1696
1697	int rc = qedFreeImage(pImage, fDelete);
1698	RTMemFree(pImage);
1699
1700	LogFlowFunc(("returns %Rrc\n", rc));
1701	return rc;
1702	}
1703
1704	/** @copydoc VDIMAGEBACKEND::pfnRead */
1705	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1706	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1707	{
1708	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1709	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1710	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1711	uint32_t offCluster = 0;
1712	uint32_t idxL1 = 0;
1713	uint32_t idxL2 = 0;
1714	uint64_t offFile = 0;
1715
1716	AssertPtr(pImage);
1717	Assert(uOffset % 512 == 0);
1718	Assert(cbToRead % 512 == 0);
1719	AssertPtrReturn(pIoCtx, VERR_INVALID_POINTER);
1720	AssertReturn(cbToRead, VERR_INVALID_PARAMETER);
1721	AssertReturn(uOffset + cbToRead <= pImage->cbSize, VERR_INVALID_PARAMETER);
1722
1723	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1724
1725	/* Clip read size to remain in the cluster. */
1726	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1727
1728	/* Get offset in image. */
1729	int rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1730	if (RT_SUCCESS(rc))
1731	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1732	pIoCtx, cbToRead);
1733
1734	if ( ( RT_SUCCESS(rc)
1735	\|\| rc == VERR_VD_BLOCK_FREE
1736	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1737	&& pcbActuallyRead)
1738	*pcbActuallyRead = cbToRead;
1739
1740	LogFlowFunc(("returns %Rrc\n", rc));
1741	return rc;
1742	}
1743
1744	/** @copydoc VDIMAGEBACKEND::pfnWrite */
1745	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1746	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1747	size_t *pcbPostRead, unsigned fWrite)
1748	{
1749	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1750	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1751	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1752	uint32_t offCluster = 0;
1753	uint32_t idxL1 = 0;
1754	uint32_t idxL2 = 0;
1755	uint64_t offImage = 0;
1756	int rc = VINF_SUCCESS;
1757
1758	AssertPtr(pImage);
1759	Assert(!(uOffset % 512));
1760	Assert(!(cbToWrite % 512));
1761	AssertPtrReturn(pIoCtx, VERR_INVALID_POINTER);
1762	AssertReturn(cbToWrite, VERR_INVALID_PARAMETER);
1763	AssertReturn(uOffset + cbToWrite <= pImage->cbSize, VERR_INVALID_PARAMETER);
1764
1765	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1766	{
1767	/* Convert offset to L1, L2 index and cluster offset. */
1768	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1769
1770	/* Clip write size to remain in the cluster. */
1771	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1772	Assert(!(cbToWrite % 512));
1773
1774	/* Get offset in image. */
1775	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1776	if (RT_SUCCESS(rc))
1777	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1778	offImage, pIoCtx, cbToWrite, NULL, NULL);
1779	else if (rc == VERR_VD_BLOCK_FREE)
1780	{
1781	if ( cbToWrite == pImage->cbCluster
1782	&& !(fWrite & VD_WRITE_NO_ALLOC))
1783	{
1784	PQEDL2CACHEENTRY pL2Entry = NULL;
1785
1786	/* Full cluster write to previously unallocated cluster.
1787	* Allocate cluster and write data. */
1788	Assert(!offCluster);
1789
1790	do
1791	{
1792	/* Check if we have to allocate a new cluster for L2 tables. */
1793	if (!pImage->paL1Table[idxL1])
1794	{
1795	uint64_t offL2Tbl;
1796	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1797
1798	/* Allocate new async cluster allocation state. */
1799	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1800	if (RT_UNLIKELY(!pL2ClusterAlloc))
1801	{
1802	rc = VERR_NO_MEMORY;
1803	break;
1804	}
1805
1806	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1807	if (!pL2Entry)
1808	{
1809	rc = VERR_NO_MEMORY;
1810	RTMemFree(pL2ClusterAlloc);
1811	break;
1812	}
1813
1814	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1815	pL2Entry->offL2Tbl = offL2Tbl;
1816	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1817
1818	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1819	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1820	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1821	pL2ClusterAlloc->idxL1 = idxL1;
1822	pL2ClusterAlloc->idxL2 = idxL2;
1823	pL2ClusterAlloc->cbToWrite = cbToWrite;
1824	pL2ClusterAlloc->pL2Entry = pL2Entry;
1825
1826	pImage->pL2TblAlloc = pL2Entry;
1827
1828	LogFlowFunc(("Allocating new L2 table at cluster offset %llu\n", offL2Tbl));
1829
1830	/*
1831	* Write the L2 table first and link to the L1 table afterwards.
1832	* If something unexpected happens the worst case which can happen
1833	* is a leak of some clusters.
1834	*/
1835	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1836	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1837	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1838	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1839	break;
1840	else if (RT_FAILURE(rc))
1841	{
1842	RTMemFree(pL2ClusterAlloc);
1843	qedL2TblCacheEntryFree(pImage, pL2Entry);
1844	break;
1845	}
1846
1847	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1848	}
1849	else
1850	{
1851	LogFlowFunc(("Fetching L2 table at cluster offset %llu\n", pImage->paL1Table[idxL1]));
1852
1853	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1854	&pL2Entry);
1855
1856	if (RT_SUCCESS(rc))
1857	{
1858	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1859
1860	/* Allocate new async cluster allocation state. */
1861	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1862	if (RT_UNLIKELY(!pDataClusterAlloc))
1863	{
1864	rc = VERR_NO_MEMORY;
1865	break;
1866	}
1867
1868	/* Allocate new cluster for the data. */
1869	uint64_t offData = qedClusterAllocate(pImage, 1);
1870
1871	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1872	pDataClusterAlloc->cbImageOld = offData;
1873	pDataClusterAlloc->offClusterNew = offData;
1874	pDataClusterAlloc->idxL1 = idxL1;
1875	pDataClusterAlloc->idxL2 = idxL2;
1876	pDataClusterAlloc->cbToWrite = cbToWrite;
1877	pDataClusterAlloc->pL2Entry = pL2Entry;
1878
1879	/* Write data. */
1880	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1881	offData, pIoCtx, cbToWrite,
1882	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1883	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1884	break;
1885	else if (RT_FAILURE(rc))
1886	{
1887	RTMemFree(pDataClusterAlloc);
1888	break;
1889	}
1890
1891	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1892	}
1893	}
1894
1895	} while (0);
1896
1897	*pcbPreRead = 0;
1898	*pcbPostRead = 0;
1899	}
1900	else
1901	{
1902	/* Trying to do a partial write to an unallocated cluster. Don't do
1903	* anything except letting the upper layer know what to do. */
1904	*pcbPreRead = offCluster;
1905	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
1906	}
1907	}
1908
1909	if (pcbWriteProcess)
1910	*pcbWriteProcess = cbToWrite;
1911	}
1912	else
1913	rc = VERR_VD_IMAGE_READ_ONLY;
1914
1915	LogFlowFunc(("returns %Rrc\n", rc));
1916	return rc;
1917	}
1918
1919	/** @copydoc VDIMAGEBACKEND::pfnFlush */
1920	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
1921	{
1922	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1923	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1924	int rc = VINF_SUCCESS;
1925
1926	AssertPtr(pImage);
1927	AssertPtrReturn(pIoCtx, VERR_INVALID_PARAMETER);
1928
1929	if ( pImage->pStorage
1930	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
1931	{
1932	QedHeader Header;
1933
1934	Assert(!(pImage->cbTable % pImage->cbCluster));
1935	rc = qedTblWrite(pImage, pIoCtx, pImage->offL1Table, pImage->paL1Table,
1936	NULL, NULL);
1937	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1938	{
1939	/* Write header. */
1940	qedHdrConvertFromHostEndianess(pImage, &Header);
1941	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1942	0, &Header, sizeof(Header),
1943	pIoCtx, NULL, NULL);
1944	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1945	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
1946	pIoCtx, NULL, NULL);
1947	}
1948	}
1949
1950	LogFlowFunc(("returns %Rrc\n", rc));
1951	return rc;
1952	}
1953
1954	/** @copydoc VDIMAGEBACKEND::pfnGetVersion */
1955	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
1956	{
1957	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1958	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1959
1960	AssertPtrReturn(pImage, 0);
1961
1962	return 1;
1963	}
1964
1965	/** @copydoc VDIMAGEBACKEND::pfnGetFileSize */
1966	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
1967	{
1968	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
1969	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1970	uint64_t cb = 0;
1971
1972	AssertPtrReturn(pImage, 0);
1973
1974	uint64_t cbFile;
1975	if (pImage->pStorage)
1976	{
1977	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1978	if (RT_SUCCESS(rc))
1979	cb += cbFile;
1980	}
1981
1982	LogFlowFunc(("returns %lld\n", cb));
1983	return cb;
1984	}
1985
1986	/** @copydoc VDIMAGEBACKEND::pfnGetPCHSGeometry */
1987	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
1988	PVDGEOMETRY pPCHSGeometry)
1989	{
1990	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
1991	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1992	int rc = VINF_SUCCESS;
1993
1994	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
1995
1996	if (pImage->PCHSGeometry.cCylinders)
1997	*pPCHSGeometry = pImage->PCHSGeometry;
1998	else
1999	rc = VERR_VD_GEOMETRY_NOT_SET;
2000
2001	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2002	return rc;
2003	}
2004
2005	/** @copydoc VDIMAGEBACKEND::pfnSetPCHSGeometry */
2006	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
2007	PCVDGEOMETRY pPCHSGeometry)
2008	{
2009	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n",
2010	pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2011	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2012	int rc = VINF_SUCCESS;
2013
2014	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2015
2016	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2017	rc = VERR_VD_IMAGE_READ_ONLY;
2018	else
2019	pImage->PCHSGeometry = *pPCHSGeometry;
2020
2021	LogFlowFunc(("returns %Rrc\n", rc));
2022	return rc;
2023	}
2024
2025	/** @copydoc VDIMAGEBACKEND::pfnGetLCHSGeometry */
2026	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData, PVDGEOMETRY pLCHSGeometry)
2027	{
2028	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2029	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2030	int rc = VINF_SUCCESS;
2031
2032	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2033
2034	if (pImage->LCHSGeometry.cCylinders)
2035	*pLCHSGeometry = pImage->LCHSGeometry;
2036	else
2037	rc = VERR_VD_GEOMETRY_NOT_SET;
2038
2039	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders,
2040	pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2041	return rc;
2042	}
2043
2044	/** @copydoc VDIMAGEBACKEND::pfnSetLCHSGeometry */
2045	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData, PCVDGEOMETRY pLCHSGeometry)
2046	{
2047	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData,
2048	pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2049	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2050	int rc = VINF_SUCCESS;
2051
2052	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2053
2054	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2055	rc = VERR_VD_IMAGE_READ_ONLY;
2056	else
2057	pImage->LCHSGeometry = *pLCHSGeometry;
2058
2059	LogFlowFunc(("returns %Rrc\n", rc));
2060	return rc;
2061	}
2062
2063	/** @copydoc VDIMAGEBACKEND::pfnQueryRegions */
2064	static DECLCALLBACK(int) qedQueryRegions(void pBackendData, PCVDREGIONLIST ppRegionList)
2065	{
2066	LogFlowFunc(("pBackendData=%#p ppRegionList=%#p\n", pBackendData, ppRegionList));
2067	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2068
2069	AssertPtrReturn(pThis, VERR_VD_NOT_OPENED);
2070
2071	*ppRegionList = &pThis->RegionList;
2072	LogFlowFunc(("returns %Rrc\n", VINF_SUCCESS));
2073	return VINF_SUCCESS;
2074	}
2075
2076	/** @copydoc VDIMAGEBACKEND::pfnRegionListRelease */
2077	static DECLCALLBACK(void) qedRegionListRelease(void *pBackendData, PCVDREGIONLIST pRegionList)
2078	{
2079	RT_NOREF1(pRegionList);
2080	LogFlowFunc(("pBackendData=%#p pRegionList=%#p\n", pBackendData, pRegionList));
2081	PQEDIMAGE pThis = (PQEDIMAGE)pBackendData;
2082	AssertPtr(pThis); RT_NOREF(pThis);
2083
2084	/* Nothing to do here. */
2085	}
2086
2087	/** @copydoc VDIMAGEBACKEND::pfnGetImageFlags */
2088	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2089	{
2090	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2091	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2092
2093	AssertPtrReturn(pImage, 0);
2094
2095	LogFlowFunc(("returns %#x\n", pImage->uImageFlags));
2096	return pImage->uImageFlags;
2097	}
2098
2099	/** @copydoc VDIMAGEBACKEND::pfnGetOpenFlags */
2100	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2101	{
2102	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2103	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2104
2105	AssertPtrReturn(pImage, 0);
2106
2107	LogFlowFunc(("returns %#x\n", pImage->uOpenFlags));
2108	return pImage->uOpenFlags;
2109	}
2110
2111	/** @copydoc VDIMAGEBACKEND::pfnSetOpenFlags */
2112	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2113	{
2114	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2115	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2116	int rc = VINF_SUCCESS;
2117
2118	/* Image must be opened and the new flags must be valid. */
2119	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2120	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SHAREABLE
2121	\| VD_OPEN_FLAGS_SEQUENTIAL \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2122	rc = VERR_INVALID_PARAMETER;
2123	else
2124	{
2125	/* Implement this operation via reopening the image. */
2126	rc = qedFreeImage(pImage, false);
2127	if (RT_SUCCESS(rc))
2128	rc = qedOpenImage(pImage, uOpenFlags);
2129	}
2130
2131	LogFlowFunc(("returns %Rrc\n", rc));
2132	return rc;
2133	}
2134
2135	/** @copydoc VDIMAGEBACKEND::pfnGetComment */
2136	VD_BACKEND_CALLBACK_GET_COMMENT_DEF_NOT_SUPPORTED(qedGetComment);
2137
2138	/** @copydoc VDIMAGEBACKEND::pfnSetComment */
2139	VD_BACKEND_CALLBACK_SET_COMMENT_DEF_NOT_SUPPORTED(qedSetComment, PQEDIMAGE);
2140
2141	/** @copydoc VDIMAGEBACKEND::pfnGetUuid */
2142	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetUuid);
2143
2144	/** @copydoc VDIMAGEBACKEND::pfnSetUuid */
2145	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetUuid, PQEDIMAGE);
2146
2147	/** @copydoc VDIMAGEBACKEND::pfnGetModificationUuid */
2148	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetModificationUuid);
2149
2150	/** @copydoc VDIMAGEBACKEND::pfnSetModificationUuid */
2151	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetModificationUuid, PQEDIMAGE);
2152
2153	/** @copydoc VDIMAGEBACKEND::pfnGetParentUuid */
2154	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetParentUuid);
2155
2156	/** @copydoc VDIMAGEBACKEND::pfnSetParentUuid */
2157	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetParentUuid, PQEDIMAGE);
2158
2159	/** @copydoc VDIMAGEBACKEND::pfnGetParentModificationUuid */
2160	VD_BACKEND_CALLBACK_GET_UUID_DEF_NOT_SUPPORTED(qedGetParentModificationUuid);
2161
2162	/** @copydoc VDIMAGEBACKEND::pfnSetParentModificationUuid */
2163	VD_BACKEND_CALLBACK_SET_UUID_DEF_NOT_SUPPORTED(qedSetParentModificationUuid, PQEDIMAGE);
2164
2165	/** @copydoc VDIMAGEBACKEND::pfnDump */
2166	static DECLCALLBACK(void) qedDump(void *pBackendData)
2167	{
2168	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2169
2170	AssertPtrReturnVoid(pImage);
2171	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2172	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2173	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2174	pImage->cbSize / 512);
2175	}
2176
2177	/** @copydoc VDIMAGEBACKEND::pfnGetParentFilename */
2178	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2179	{
2180	int rc = VINF_SUCCESS;
2181	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2182
2183	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2184
2185	if (pImage->pszBackingFilename)
2186	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2187	else
2188	rc = VERR_NOT_SUPPORTED;
2189
2190	LogFlowFunc(("returns %Rrc\n", rc));
2191	return rc;
2192	}
2193
2194	/** @copydoc VDIMAGEBACKEND::pfnSetParentFilename */
2195	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2196	{
2197	int rc = VINF_SUCCESS;
2198	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2199
2200	AssertPtrReturn(pImage, VERR_VD_NOT_OPENED);
2201
2202	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2203	rc = VERR_VD_IMAGE_READ_ONLY;
2204	else if ( pImage->pszBackingFilename
2205	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2206	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2207	else
2208	{
2209	if (pImage->pszBackingFilename)
2210	RTStrFree(pImage->pszBackingFilename);
2211	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2212	if (!pImage->pszBackingFilename)
2213	rc = VERR_NO_STR_MEMORY;
2214	else
2215	{
2216	if (!pImage->offBackingFilename)
2217	{
2218	/* Allocate new cluster. */
2219	uint64_t offData = qedClusterAllocate(pImage, 1);
2220
2221	Assert((offData & UINT32_MAX) == offData);
2222	pImage->offBackingFilename = (uint32_t)offData;
2223	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2224	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2225	offData + pImage->cbCluster);
2226	}
2227
2228	if (RT_SUCCESS(rc))
2229	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2230	pImage->offBackingFilename,
2231	pImage->pszBackingFilename,
2232	strlen(pImage->pszBackingFilename));
2233	}
2234	}
2235
2236	LogFlowFunc(("returns %Rrc\n", rc));
2237	return rc;
2238	}
2239
2240	/** @copydoc VDIMAGEBACKEND::pfnResize */
2241	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2242	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2243	unsigned uPercentStart, unsigned uPercentSpan,
2244	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2245	PVDINTERFACE pVDIfsOperation)
2246	{
2247	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2248	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2249	int rc = VINF_SUCCESS;
2250
2251	/* Making the image smaller is not supported at the moment. */
2252	if (cbSize < pImage->cbSize)
2253	rc = VERR_NOT_SUPPORTED;
2254	else if (cbSize > pImage->cbSize)
2255	{
2256	/*
2257	* It is enough to just update the size field in the header to complete
2258	* growing. With the default cluster and table sizes the image can be expanded
2259	* to 64TB without overflowing the L1 and L2 tables making block relocation
2260	* superfluous.
2261	* @todo: The rare case where block relocation is still required (non default
2262	* table and/or cluster size or images with more than 64TB) is not
2263	* implemented yet and resizing such an image will fail with an error.
2264	*/
2265	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2266	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2267	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2268	pImage->pszFilename);
2269	else
2270	{
2271	uint64_t cbSizeOld = pImage->cbSize;
2272
2273	pImage->cbSize = cbSize;
2274	rc = qedFlushImage(pImage);
2275	if (RT_FAILURE(rc))
2276	{
2277	pImage->cbSize = cbSizeOld; /* Restore */
2278
2279	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2280	pImage->pszFilename);
2281	}
2282	}
2283	}
2284	/* Same size doesn't change the image at all. */
2285
2286	LogFlowFunc(("returns %Rrc\n", rc));
2287	return rc;
2288	}
2289
2290
2291	const VDIMAGEBACKEND g_QedBackend =
2292	{
2293	/* u32Version */
2294	VD_IMGBACKEND_VERSION,
2295	/* pszBackendName */
2296	"QED",
2297	/* uBackendCaps */
2298	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2299	/* paFileExtensions */
2300	s_aQedFileExtensions,
2301	/* paConfigInfo */
2302	NULL,
2303	/* pfnProbe */
2304	qedProbe,
2305	/* pfnOpen */
2306	qedOpen,
2307	/* pfnCreate */
2308	qedCreate,
2309	/* pfnRename */
2310	qedRename,
2311	/* pfnClose */
2312	qedClose,
2313	/* pfnRead */
2314	qedRead,
2315	/* pfnWrite */
2316	qedWrite,
2317	/* pfnFlush */
2318	qedFlush,
2319	/* pfnDiscard */
2320	NULL,
2321	/* pfnGetVersion */
2322	qedGetVersion,
2323	/* pfnGetFileSize */
2324	qedGetFileSize,
2325	/* pfnGetPCHSGeometry */
2326	qedGetPCHSGeometry,
2327	/* pfnSetPCHSGeometry */
2328	qedSetPCHSGeometry,
2329	/* pfnGetLCHSGeometry */
2330	qedGetLCHSGeometry,
2331	/* pfnSetLCHSGeometry */
2332	qedSetLCHSGeometry,
2333	/* pfnQueryRegions */
2334	qedQueryRegions,
2335	/* pfnRegionListRelease */
2336	qedRegionListRelease,
2337	/* pfnGetImageFlags */
2338	qedGetImageFlags,
2339	/* pfnGetOpenFlags */
2340	qedGetOpenFlags,
2341	/* pfnSetOpenFlags */
2342	qedSetOpenFlags,
2343	/* pfnGetComment */
2344	qedGetComment,
2345	/* pfnSetComment */
2346	qedSetComment,
2347	/* pfnGetUuid */
2348	qedGetUuid,
2349	/* pfnSetUuid */
2350	qedSetUuid,
2351	/* pfnGetModificationUuid */
2352	qedGetModificationUuid,
2353	/* pfnSetModificationUuid */
2354	qedSetModificationUuid,
2355	/* pfnGetParentUuid */
2356	qedGetParentUuid,
2357	/* pfnSetParentUuid */
2358	qedSetParentUuid,
2359	/* pfnGetParentModificationUuid */
2360	qedGetParentModificationUuid,
2361	/* pfnSetParentModificationUuid */
2362	qedSetParentModificationUuid,
2363	/* pfnDump */
2364	qedDump,
2365	/* pfnGetTimestamp */
2366	NULL,
2367	/* pfnGetParentTimestamp */
2368	NULL,
2369	/* pfnSetParentTimestamp */
2370	NULL,
2371	/* pfnGetParentFilename */
2372	qedGetParentFilename,
2373	/* pfnSetParentFilename */
2374	qedSetParentFilename,
2375	/* pfnComposeLocation */
2376	genericFileComposeLocation,
2377	/* pfnComposeName */
2378	genericFileComposeName,
2379	/* pfnCompact */
2380	NULL,
2381	/* pfnResize */
2382	qedResize,
2383	/* pfnRepair */
2384	NULL,
2385	/* pfnTraverseMetadata */
2386	NULL,
2387	/* u32Version */
2388	VD_IMGBACKEND_VERSION
2389	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 100964

Download in other formats: