QED.cpp@ 56993

Last change on this file since 56993 was 56993, checked in by vboxsync, 9 years ago
Storage: Log & Assertion formatting fixes.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 88.1 KB

Line
1	/* $Id: QED.cpp 56993 2015-07-18 23:02:09Z vboxsync $ */
2	/** @file
3	* QED - QED Disk image.
4	*/
5
6	/*
7	* Copyright (C) 2011-2015 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	* Header Files *
20	*******************************************************************************/
21	#define LOG_GROUP LOG_GROUP_VD_QED
22	#include <VBox/vd-plugin.h>
23	#include <VBox/err.h>
24
25	#include <VBox/log.h>
26	#include <iprt/asm.h>
27	#include <iprt/assert.h>
28	#include <iprt/string.h>
29	#include <iprt/alloc.h>
30	#include <iprt/path.h>
31	#include <iprt/list.h>
32
33	#include "VDBackends.h"
34
35	/**
36	* The QED backend implements support for the qemu enhanced disk format (short QED)
37	* The specification for the format is available under http://wiki.qemu.org/Features/QED/Specification
38	*
39	* Missing things to implement:
40	* - compaction
41	* - resizing which requires block relocation (very rare case)
42	*/
43
44	/*******************************************************************************
45	* Structures in a QED image, little endian *
46	*******************************************************************************/
47
48	#pragma pack(1)
49	typedef struct QedHeader
50	{
51	/** Magic value. */
52	uint32_t u32Magic;
53	/** Cluster size in bytes. */
54	uint32_t u32ClusterSize;
55	/** Size of L1 and L2 tables in clusters. */
56	uint32_t u32TableSize;
57	/** size of this header structure in clusters. */
58	uint32_t u32HeaderSize;
59	/** Features used for the image. */
60	uint64_t u64FeatureFlags;
61	/** Compatibility features used for the image. */
62	uint64_t u64CompatFeatureFlags;
63	/** Self resetting feature bits. */
64	uint64_t u64AutoresetFeatureFlags;
65	/** Offset of the L1 table in bytes. */
66	uint64_t u64OffL1Table;
67	/** Logical image size as seen by the guest. */
68	uint64_t u64Size;
69	/** Offset of the backing filename in bytes. */
70	uint32_t u32OffBackingFilename;
71	/** Size of the backing filename. */
72	uint32_t u32BackingFilenameSize;
73	} QedHeader;
74	#pragma pack()
75	/** Pointer to a on disk QED header. */
76	typedef QedHeader *PQedHeader;
77
78	/** QED magic value. */
79	#define QED_MAGIC UINT32_C(0x00444551) /* QED\0 */
80	/** Cluster size minimum. */
81	#define QED_CLUSTER_SIZE_MIN RT_BIT(12)
82	/** Cluster size maximum. */
83	#define QED_CLUSTER_SIZE_MAX RT_BIT(26)
84	/** L1 and L2 Table size minimum. */
85	#define QED_TABLE_SIZE_MIN 1
86	/** L1 and L2 Table size maximum. */
87	#define QED_TABLE_SIZE_MAX 16
88
89	/** QED default cluster size when creating an image. */
90	#define QED_CLUSTER_SIZE_DEFAULT (64 * _1K)
91	/** The default table size in clusters. */
92	#define QED_TABLE_SIZE_DEFAULT 4
93
94	/** Feature flags.
95	* @{
96	*/
97	/** Image uses a backing file to provide data for unallocated clusters. */
98	#define QED_FEATURE_BACKING_FILE RT_BIT_64(0)
99	/** Image needs checking before use. */
100	#define QED_FEATURE_NEED_CHECK RT_BIT_64(1)
101	/** Don't probe for format of the backing file, treat as raw image. */
102	#define QED_FEATURE_BACKING_FILE_NO_PROBE RT_BIT_64(2)
103	/** Mask of valid features. */
104	#define QED_FEATURE_MASK (QED_FEATURE_BACKING_FILE \| QED_FEATURE_NEED_CHECK \| QED_FEATURE_BACKING_FILE_NO_PROBE)
105	/** @} */
106
107	/** Compatibility feature flags.
108	* @{
109	*/
110	/** Mask of valid compatibility features. */
111	#define QED_COMPAT_FEATURE_MASK (0)
112	/** @} */
113
114	/** Autoreset feature flags.
115	* @{
116	*/
117	/** Mask of valid autoreset features. */
118	#define QED_AUTORESET_FEATURE_MASK (0)
119	/** @} */
120
121	/*******************************************************************************
122	* Constants And Macros, Structures and Typedefs *
123	*******************************************************************************/
124
125	/**
126	* QED L2 cache entry.
127	*/
128	typedef struct QEDL2CACHEENTRY
129	{
130	/** List node for the search list. */
131	RTLISTNODE NodeSearch;
132	/** List node for the LRU list. */
133	RTLISTNODE NodeLru;
134	/** Reference counter. */
135	uint32_t cRefs;
136	/** The offset of the L2 table, used as search key. */
137	uint64_t offL2Tbl;
138	/** Pointer to the cached L2 table. */
139	uint64_t *paL2Tbl;
140	} QEDL2CACHEENTRY, *PQEDL2CACHEENTRY;
141
142	/** Maximum amount of memory the cache is allowed to use. */
143	#define QED_L2_CACHE_MEMORY_MAX (2*_1M)
144
145	/**
146	* QED image data structure.
147	*/
148	typedef struct QEDIMAGE
149	{
150	/** Image name. */
151	const char *pszFilename;
152	/** Storage handle. */
153	PVDIOSTORAGE pStorage;
154
155	/** Pointer to the per-disk VD interface list. */
156	PVDINTERFACE pVDIfsDisk;
157	/** Pointer to the per-image VD interface list. */
158	PVDINTERFACE pVDIfsImage;
159	/** Error interface. */
160	PVDINTERFACEERROR pIfError;
161	/** I/O interface. */
162	PVDINTERFACEIOINT pIfIo;
163
164	/** Open flags passed by VBoxHD layer. */
165	unsigned uOpenFlags;
166	/** Image flags defined during creation or determined during open. */
167	unsigned uImageFlags;
168	/** Total size of the image. */
169	uint64_t cbSize;
170	/** Physical geometry of this image. */
171	VDGEOMETRY PCHSGeometry;
172	/** Logical geometry of this image. */
173	VDGEOMETRY LCHSGeometry;
174
175	/** Filename of the backing file if any. */
176	char *pszBackingFilename;
177	/** Offset of the filename in the image. */
178	uint32_t offBackingFilename;
179	/** Size of the backing filename excluding \0. */
180	uint32_t cbBackingFilename;
181
182	/** Size of the image, multiple of clusters. */
183	uint64_t cbImage;
184	/** Cluster size in bytes. */
185	uint32_t cbCluster;
186	/** Number of entries in the L1 and L2 table. */
187	uint32_t cTableEntries;
188	/** Size of an L1 or L2 table rounded to the next cluster size. */
189	uint32_t cbTable;
190	/** Pointer to the L1 table. */
191	uint64_t *paL1Table;
192	/** Offset of the L1 table. */
193	uint64_t offL1Table;
194
195	/** Offset mask for a cluster. */
196	uint64_t fOffsetMask;
197	/** L1 table mask to get the L1 index. */
198	uint64_t fL1Mask;
199	/** Number of bits to shift to get the L1 index. */
200	uint32_t cL1Shift;
201	/** L2 table mask to get the L2 index. */
202	uint64_t fL2Mask;
203	/** Number of bits to shift to get the L2 index. */
204	uint32_t cL2Shift;
205
206	/** Memory occupied by the L2 table cache. */
207	size_t cbL2Cache;
208	/** The sorted L2 entry list used for searching. */
209	RTLISTNODE ListSearch;
210	/** The LRU L2 entry list used for eviction. */
211	RTLISTNODE ListLru;
212
213	} QEDIMAGE, *PQEDIMAGE;
214
215	/**
216	* State of the async cluster allocation.
217	*/
218	typedef enum QEDCLUSTERASYNCALLOCSTATE
219	{
220	/** Invalid. */
221	QEDCLUSTERASYNCALLOCSTATE_INVALID = 0,
222	/** L2 table allocation. */
223	QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC,
224	/** Link L2 table into L1. */
225	QEDCLUSTERASYNCALLOCSTATE_L2_LINK,
226	/** Allocate user data cluster. */
227	QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC,
228	/** Link user data cluster. */
229	QEDCLUSTERASYNCALLOCSTATE_USER_LINK,
230	/** 32bit blowup. */
231	QEDCLUSTERASYNCALLOCSTATE_32BIT_HACK = 0x7fffffff
232	} QEDCLUSTERASYNCALLOCSTATE, *PQEDCLUSTERASYNCALLOCSTATE;
233
234	/**
235	* Data needed to track async cluster allocation.
236	*/
237	typedef struct QEDCLUSTERASYNCALLOC
238	{
239	/** The state of the cluster allocation. */
240	QEDCLUSTERASYNCALLOCSTATE enmAllocState;
241	/** Old image size to rollback in case of an error. */
242	uint64_t cbImageOld;
243	/** L1 index to link if any. */
244	uint32_t idxL1;
245	/** L2 index to link, required in any case. */
246	uint32_t idxL2;
247	/** Start offset of the allocated cluster. */
248	uint64_t offClusterNew;
249	/** L2 cache entry if a L2 table is allocated. */
250	PQEDL2CACHEENTRY pL2Entry;
251	/** Number of bytes to write. */
252	size_t cbToWrite;
253	} QEDCLUSTERASYNCALLOC, *PQEDCLUSTERASYNCALLOC;
254
255	/*******************************************************************************
256	* Static Variables *
257	*******************************************************************************/
258
259	/** NULL-terminated array of supported file extensions. */
260	static const VDFILEEXTENSION s_aQedFileExtensions[] =
261	{
262	{"qed", VDTYPE_HDD},
263	{NULL, VDTYPE_INVALID}
264	};
265
266	/*******************************************************************************
267	* Internal Functions *
268	*******************************************************************************/
269
270	/**
271	* Converts the image header to the host endianess and performs basic checks.
272	*
273	* @returns Whether the given header is valid or not.
274	* @param pHeader Pointer to the header to convert.
275	*/
276	static bool qedHdrConvertToHostEndianess(PQedHeader pHeader)
277	{
278	pHeader->u32Magic = RT_LE2H_U32(pHeader->u32Magic);
279	pHeader->u32ClusterSize = RT_LE2H_U32(pHeader->u32ClusterSize);
280	pHeader->u32TableSize = RT_LE2H_U32(pHeader->u32TableSize);
281	pHeader->u32HeaderSize = RT_LE2H_U32(pHeader->u32HeaderSize);
282	pHeader->u64FeatureFlags = RT_LE2H_U64(pHeader->u64FeatureFlags);
283	pHeader->u64CompatFeatureFlags = RT_LE2H_U64(pHeader->u64CompatFeatureFlags);
284	pHeader->u64AutoresetFeatureFlags = RT_LE2H_U64(pHeader->u64AutoresetFeatureFlags);
285	pHeader->u64OffL1Table = RT_LE2H_U64(pHeader->u64OffL1Table);
286	pHeader->u64Size = RT_LE2H_U64(pHeader->u64Size);
287	pHeader->u32OffBackingFilename = RT_LE2H_U32(pHeader->u32OffBackingFilename);
288	pHeader->u32BackingFilenameSize = RT_LE2H_U32(pHeader->u32BackingFilenameSize);
289
290	if (RT_UNLIKELY(pHeader->u32Magic != QED_MAGIC))
291	return false;
292	if (RT_UNLIKELY( pHeader->u32ClusterSize < QED_CLUSTER_SIZE_MIN
293	\|\| pHeader->u32ClusterSize > QED_CLUSTER_SIZE_MAX))
294	return false;
295	if (RT_UNLIKELY( pHeader->u32TableSize < QED_TABLE_SIZE_MIN
296	\|\| pHeader->u32TableSize > QED_TABLE_SIZE_MAX))
297	return false;
298	if (RT_UNLIKELY(pHeader->u64Size % 512 != 0))
299	return false;
300
301	return true;
302	}
303
304	/**
305	* Creates a QED header from the given image state.
306	*
307	* @returns nothing.
308	* @param pImage Image instance data.
309	* @param pHeader Pointer to the header to convert.
310	*/
311	static void qedHdrConvertFromHostEndianess(PQEDIMAGE pImage, PQedHeader pHeader)
312	{
313	pHeader->u32Magic = RT_H2LE_U32(QED_MAGIC);
314	pHeader->u32ClusterSize = RT_H2LE_U32(pImage->cbCluster);
315	pHeader->u32TableSize = RT_H2LE_U32(pImage->cbTable / pImage->cbCluster);
316	pHeader->u32HeaderSize = RT_H2LE_U32(1);
317	pHeader->u64FeatureFlags = RT_H2LE_U64(pImage->pszBackingFilename ? QED_FEATURE_BACKING_FILE : UINT64_C(0));
318	pHeader->u64CompatFeatureFlags = RT_H2LE_U64(UINT64_C(0));
319	pHeader->u64AutoresetFeatureFlags = RT_H2LE_U64(UINT64_C(0));
320	pHeader->u64OffL1Table = RT_H2LE_U64(pImage->offL1Table);
321	pHeader->u64Size = RT_H2LE_U64(pImage->cbSize);
322	pHeader->u32OffBackingFilename = RT_H2LE_U32(pImage->offBackingFilename);
323	pHeader->u32BackingFilenameSize = RT_H2LE_U32(pImage->cbBackingFilename);
324	}
325
326	/**
327	* Convert table entries from little endian to host endianess.
328	*
329	* @returns nothing.
330	* @param paTbl Pointer to the table.
331	* @param cEntries Number of entries in the table.
332	*/
333	static void qedTableConvertToHostEndianess(uint64_t *paTbl, uint32_t cEntries)
334	{
335	while(cEntries-- > 0)
336	{
337	paTbl = RT_LE2H_U64(paTbl);
338	paTbl++;
339	}
340	}
341
342	/**
343	* Convert table entries from host to little endian format.
344	*
345	* @returns nothing.
346	* @param paTblImg Pointer to the table which will store the little endian table.
347	* @param paTbl The source table to convert.
348	* @param cEntries Number of entries in the table.
349	*/
350	static void qedTableConvertFromHostEndianess(uint64_t paTblImg, uint64_t paTbl,
351	uint32_t cEntries)
352	{
353	while(cEntries-- > 0)
354	{
355	paTblImg = RT_H2LE_U64(paTbl);
356	paTbl++;
357	paTblImg++;
358	}
359	}
360
361	/**
362	* Creates the L2 table cache.
363	*
364	* @returns VBox status code.
365	* @param pImage The image instance data.
366	*/
367	static int qedL2TblCacheCreate(PQEDIMAGE pImage)
368	{
369	pImage->cbL2Cache = 0;
370	RTListInit(&pImage->ListSearch);
371	RTListInit(&pImage->ListLru);
372
373	return VINF_SUCCESS;
374	}
375
376	/**
377	* Destroys the L2 table cache.
378	*
379	* @returns nothing.
380	* @param pImage The image instance data.
381	*/
382	static void qedL2TblCacheDestroy(PQEDIMAGE pImage)
383	{
384	PQEDL2CACHEENTRY pL2Entry = NULL;
385	PQEDL2CACHEENTRY pL2Next = NULL;
386
387	RTListForEachSafe(&pImage->ListSearch, pL2Entry, pL2Next, QEDL2CACHEENTRY, NodeSearch)
388	{
389	Assert(!pL2Entry->cRefs);
390
391	RTListNodeRemove(&pL2Entry->NodeSearch);
392	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
393	RTMemFree(pL2Entry);
394	}
395
396	pImage->cbL2Cache = 0;
397	RTListInit(&pImage->ListSearch);
398	RTListInit(&pImage->ListLru);
399	}
400
401	/**
402	* Returns the L2 table matching the given offset or NULL if none could be found.
403	*
404	* @returns Pointer to the L2 table cache entry or NULL.
405	* @param pImage The image instance data.
406	* @param offL2Tbl Offset of the L2 table to search for.
407	*/
408	static PQEDL2CACHEENTRY qedL2TblCacheRetain(PQEDIMAGE pImage, uint64_t offL2Tbl)
409	{
410	PQEDL2CACHEENTRY pL2Entry = NULL;
411
412	RTListForEach(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch)
413	{
414	if (pL2Entry->offL2Tbl == offL2Tbl)
415	break;
416	}
417
418	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
419	{
420	/* Update LRU list. */
421	RTListNodeRemove(&pL2Entry->NodeLru);
422	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
423	pL2Entry->cRefs++;
424	return pL2Entry;
425	}
426	else
427	return NULL;
428	}
429
430	/**
431	* Releases a L2 table cache entry.
432	*
433	* @returns nothing.
434	* @param pL2Entry The L2 cache entry.
435	*/
436	static void qedL2TblCacheEntryRelease(PQEDL2CACHEENTRY pL2Entry)
437	{
438	Assert(pL2Entry->cRefs > 0);
439	pL2Entry->cRefs--;
440	}
441
442	/**
443	* Allocates a new L2 table from the cache evicting old entries if required.
444	*
445	* @returns Pointer to the L2 cache entry or NULL.
446	* @param pImage The image instance data.
447	*/
448	static PQEDL2CACHEENTRY qedL2TblCacheEntryAlloc(PQEDIMAGE pImage)
449	{
450	PQEDL2CACHEENTRY pL2Entry = NULL;
451	int rc = VINF_SUCCESS;
452
453	if (pImage->cbL2Cache + pImage->cbTable <= QED_L2_CACHE_MEMORY_MAX)
454	{
455	/* Add a new entry. */
456	pL2Entry = (PQEDL2CACHEENTRY)RTMemAllocZ(sizeof(QEDL2CACHEENTRY));
457	if (pL2Entry)
458	{
459	pL2Entry->paL2Tbl = (uint64_t *)RTMemPageAllocZ(pImage->cbTable);
460	if (RT_UNLIKELY(!pL2Entry->paL2Tbl))
461	{
462	RTMemFree(pL2Entry);
463	pL2Entry = NULL;
464	}
465	else
466	{
467	pL2Entry->cRefs = 1;
468	pImage->cbL2Cache += pImage->cbTable;
469	}
470	}
471	}
472	else
473	{
474	/* Evict the last not in use entry and use it */
475	Assert(!RTListIsEmpty(&pImage->ListLru));
476
477	RTListForEachReverse(&pImage->ListLru, pL2Entry, QEDL2CACHEENTRY, NodeLru)
478	{
479	if (!pL2Entry->cRefs)
480	break;
481	}
482
483	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
484	{
485	RTListNodeRemove(&pL2Entry->NodeSearch);
486	RTListNodeRemove(&pL2Entry->NodeLru);
487	pL2Entry->offL2Tbl = 0;
488	pL2Entry->cRefs = 1;
489	}
490	else
491	pL2Entry = NULL;
492	}
493
494	return pL2Entry;
495	}
496
497	/**
498	* Frees a L2 table cache entry.
499	*
500	* @returns nothing.
501	* @param pImage The image instance data.
502	* @param pL2Entry The L2 cache entry to free.
503	*/
504	static void qedL2TblCacheEntryFree(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
505	{
506	Assert(!pL2Entry->cRefs);
507	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
508	RTMemFree(pL2Entry);
509
510	pImage->cbL2Cache -= pImage->cbTable;
511	}
512
513	/**
514	* Inserts an entry in the L2 table cache.
515	*
516	* @returns nothing.
517	* @param pImage The image instance data.
518	* @param pL2Entry The L2 cache entry to insert.
519	*/
520	static void qedL2TblCacheEntryInsert(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
521	{
522	PQEDL2CACHEENTRY pIt = NULL;
523
524	Assert(pL2Entry->offL2Tbl > 0);
525
526	/* Insert at the top of the LRU list. */
527	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
528
529	if (RTListIsEmpty(&pImage->ListSearch))
530	{
531	RTListAppend(&pImage->ListSearch, &pL2Entry->NodeSearch);
532	}
533	else
534	{
535	/* Insert into search list. */
536	pIt = RTListGetFirst(&pImage->ListSearch, QEDL2CACHEENTRY, NodeSearch);
537	if (pIt->offL2Tbl > pL2Entry->offL2Tbl)
538	RTListPrepend(&pImage->ListSearch, &pL2Entry->NodeSearch);
539	else
540	{
541	bool fInserted = false;
542
543	RTListForEach(&pImage->ListSearch, pIt, QEDL2CACHEENTRY, NodeSearch)
544	{
545	Assert(pIt->offL2Tbl != pL2Entry->offL2Tbl);
546	if (pIt->offL2Tbl < pL2Entry->offL2Tbl)
547	{
548	RTListNodeInsertAfter(&pIt->NodeSearch, &pL2Entry->NodeSearch);
549	fInserted = true;
550	break;
551	}
552	}
553	Assert(fInserted);
554	}
555	}
556	}
557
558	/**
559	* Fetches the L2 from the given offset trying the LRU cache first and
560	* reading it from the image after a cache miss.
561	*
562	* @returns VBox status code.
563	* @param pImage Image instance data.
564	* @param offL2Tbl The offset of the L2 table in the image.
565	* @param ppL2Entry Where to store the L2 table on success.
566	*/
567	static int qedL2TblCacheFetch(PQEDIMAGE pImage, uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
568	{
569	int rc = VINF_SUCCESS;
570
571	LogFlowFunc(("pImage=%#p offL2Tbl=%llu ppL2Entry=%#p\n", pImage, offL2Tbl, ppL2Entry));
572
573	/* Try to fetch the L2 table from the cache first. */
574	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
575	if (!pL2Entry)
576	{
577	LogFlowFunc(("Reading L2 table from image\n"));
578	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
579
580	if (pL2Entry)
581	{
582	/* Read from the image. */
583	pL2Entry->offL2Tbl = offL2Tbl;
584	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, offL2Tbl,
585	pL2Entry->paL2Tbl, pImage->cbTable);
586	if (RT_SUCCESS(rc))
587	{
588	#if defined(RT_BIG_ENDIAN)
589	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
590	#endif
591	qedL2TblCacheEntryInsert(pImage, pL2Entry);
592	}
593	else
594	{
595	qedL2TblCacheEntryRelease(pL2Entry);
596	qedL2TblCacheEntryFree(pImage, pL2Entry);
597	}
598	}
599	else
600	rc = VERR_NO_MEMORY;
601	}
602
603	if (RT_SUCCESS(rc))
604	*ppL2Entry = pL2Entry;
605
606	LogFlowFunc(("returns rc=%Rrc\n", rc));
607	return rc;
608	}
609
610	/**
611	* Fetches the L2 from the given offset trying the LRU cache first and
612	* reading it from the image after a cache miss - version for async I/O.
613	*
614	* @returns VBox status code.
615	* @param pImage Image instance data.
616	* @param pIoCtx The I/O context.
617	* @param offL2Tbl The offset of the L2 table in the image.
618	* @param ppL2Entry Where to store the L2 table on success.
619	*/
620	static int qedL2TblCacheFetchAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
621	uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
622	{
623	int rc = VINF_SUCCESS;
624
625	/* Try to fetch the L2 table from the cache first. */
626	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
627	if (!pL2Entry)
628	{
629	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
630
631	if (pL2Entry)
632	{
633	/* Read from the image. */
634	PVDMETAXFER pMetaXfer;
635
636	pL2Entry->offL2Tbl = offL2Tbl;
637	rc = vdIfIoIntFileReadMeta(pImage->pIfIo, pImage->pStorage,
638	offL2Tbl, pL2Entry->paL2Tbl,
639	pImage->cbTable, pIoCtx,
640	&pMetaXfer, NULL, NULL);
641	if (RT_SUCCESS(rc))
642	{
643	vdIfIoIntMetaXferRelease(pImage->pIfIo, pMetaXfer);
644	#if defined(RT_BIG_ENDIAN)
645	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
646	#endif
647	qedL2TblCacheEntryInsert(pImage, pL2Entry);
648	}
649	else
650	{
651	qedL2TblCacheEntryRelease(pL2Entry);
652	qedL2TblCacheEntryFree(pImage, pL2Entry);
653	}
654	}
655	else
656	rc = VERR_NO_MEMORY;
657	}
658
659	if (RT_SUCCESS(rc))
660	*ppL2Entry = pL2Entry;
661
662	return rc;
663	}
664
665	/**
666	* Return power of 2 or 0 if num error.
667	*
668	* @returns The power of 2 or 0 if the given number is not a power of 2.
669	* @param u32 The number.
670	*/
671	static uint32_t qedGetPowerOfTwo(uint32_t u32)
672	{
673	if (u32 == 0)
674	return 0;
675	uint32_t uPower2 = 0;
676	while ((u32 & 1) == 0)
677	{
678	u32 >>= 1;
679	uPower2++;
680	}
681	return u32 == 1 ? uPower2 : 0;
682	}
683
684	/**
685	* Sets the L1, L2 and offset bitmasks and L1 and L2 bit shift members.
686	*
687	* @returns nothing.
688	* @param pImage The image instance data.
689	*/
690	static void qedTableMasksInit(PQEDIMAGE pImage)
691	{
692	uint32_t cClusterBits, cTableBits;
693
694	cClusterBits = qedGetPowerOfTwo(pImage->cbCluster);
695	cTableBits = qedGetPowerOfTwo(pImage->cTableEntries);
696
697	Assert(cClusterBits + 2 * cTableBits <= 64);
698
699	pImage->fOffsetMask = ((uint64_t)pImage->cbCluster - 1);
700	pImage->fL2Mask = ((uint64_t)pImage->cTableEntries - 1) << cClusterBits;
701	pImage->cL2Shift = cClusterBits;
702	pImage->fL1Mask = ((uint64_t)pImage->cTableEntries - 1) << (cClusterBits + cTableBits);
703	pImage->cL1Shift = cClusterBits + cTableBits;
704	}
705
706	/**
707	* Converts a given logical offset into the
708	*
709	* @returns nothing.
710	* @param pImage The image instance data.
711	* @param off The logical offset to convert.
712	* @param pidxL1 Where to store the index in the L1 table on success.
713	* @param pidxL2 Where to store the index in the L2 table on success.
714	* @param poffCluster Where to store the offset in the cluster on success.
715	*/
716	DECLINLINE(void) qedConvertLogicalOffset(PQEDIMAGE pImage, uint64_t off, uint32_t *pidxL1,
717	uint32_t pidxL2, uint32_t poffCluster)
718	{
719	AssertPtr(pidxL1);
720	AssertPtr(pidxL2);
721	AssertPtr(poffCluster);
722
723	*poffCluster = off & pImage->fOffsetMask;
724	*pidxL1 = (off & pImage->fL1Mask) >> pImage->cL1Shift;
725	*pidxL2 = (off & pImage->fL2Mask) >> pImage->cL2Shift;
726	}
727
728	/**
729	* Converts Cluster size to a byte size.
730	*
731	* @returns Number of bytes derived from the given number of clusters.
732	* @param pImage The image instance data.
733	* @param cClusters The clusters to convert.
734	*/
735	DECLINLINE(uint64_t) qedCluster2Byte(PQEDIMAGE pImage, uint64_t cClusters)
736	{
737	return cClusters * pImage->cbCluster;
738	}
739
740	/**
741	* Converts number of bytes to cluster size rounding to the next cluster.
742	*
743	* @returns Number of bytes derived from the given number of clusters.
744	* @param pImage The image instance data.
745	* @param cb Number of bytes to convert.
746	*/
747	DECLINLINE(uint64_t) qedByte2Cluster(PQEDIMAGE pImage, uint64_t cb)
748	{
749	return cb / pImage->cbCluster + (cb % pImage->cbCluster ? 1 : 0);
750	}
751
752	/**
753	* Allocates a new cluster in the image.
754	*
755	* @returns The start offset of the new cluster in the image.
756	* @param pImage The image instance data.
757	* @param cCLusters Number of clusters to allocate.
758	*/
759	DECLINLINE(uint64_t) qedClusterAllocate(PQEDIMAGE pImage, uint32_t cClusters)
760	{
761	uint64_t offCluster;
762
763	offCluster = pImage->cbImage;
764	pImage->cbImage += cClusters*pImage->cbCluster;
765
766	return offCluster;
767	}
768
769	/**
770	* Returns the real image offset for a given cluster or an error if the cluster is not
771	* yet allocated.
772	*
773	* @returns VBox status code.
774	* VERR_VD_BLOCK_FREE if the cluster is not yet allocated.
775	* @param pImage The image instance data.
776	* @param pIoCtx The I/O context.
777	* @param idxL1 The L1 index.
778	* @param idxL2 The L2 index.
779	* @param offCluster Offset inside the cluster.
780	* @param poffImage Where to store the image offset on success;
781	*/
782	static int qedConvertToImageOffset(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
783	uint32_t idxL1, uint32_t idxL2,
784	uint32_t offCluster, uint64_t *poffImage)
785	{
786	int rc = VERR_VD_BLOCK_FREE;
787
788	AssertReturn(idxL1 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
789	AssertReturn(idxL2 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
790
791	if (pImage->paL1Table[idxL1])
792	{
793	PQEDL2CACHEENTRY pL2Entry;
794
795	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
796	&pL2Entry);
797	if (RT_SUCCESS(rc))
798	{
799	/* Get real file offset. */
800	if (pL2Entry->paL2Tbl[idxL2])
801	*poffImage = pL2Entry->paL2Tbl[idxL2] + offCluster;
802	else
803	rc = VERR_VD_BLOCK_FREE;
804
805	qedL2TblCacheEntryRelease(pL2Entry);
806	}
807	}
808
809	return rc;
810	}
811
812
813	/**
814	* Internal. Flush image data to disk.
815	*/
816	static int qedFlushImage(PQEDIMAGE pImage)
817	{
818	int rc = VINF_SUCCESS;
819
820	if ( pImage->pStorage
821	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
822	{
823	QedHeader Header;
824
825	Assert(!(pImage->cbTable % pImage->cbCluster));
826	#if defined(RT_BIG_ENDIAN)
827	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
828	if (paL1TblImg)
829	{
830	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
831	pImage->cTableEntries);
832	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
833	pImage->offL1Table, paL1TblImg,
834	pImage->cbTable);
835	RTMemFree(paL1TblImg);
836	}
837	else
838	rc = VERR_NO_MEMORY;
839	#else
840	/* Write L1 table directly. */
841	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, pImage->offL1Table,
842	pImage->paL1Table, pImage->cbTable);
843	#endif
844	if (RT_SUCCESS(rc))
845	{
846	/* Write header. */
847	qedHdrConvertFromHostEndianess(pImage, &Header);
848	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, 0, &Header,
849	sizeof(Header));
850	if (RT_SUCCESS(rc))
851	rc = vdIfIoIntFileFlushSync(pImage->pIfIo, pImage->pStorage);
852	}
853	}
854
855	return rc;
856	}
857
858	/**
859	* Flush image data to disk - version for async I/O.
860	*
861	* @returns VBox status code.
862	* @param pImage The image instance data.
863	* @param pIoCtx The I/o context
864	*/
865	static int qedFlushImageAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx)
866	{
867	int rc = VINF_SUCCESS;
868
869	if ( pImage->pStorage
870	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
871	{
872	QedHeader Header;
873
874	Assert(!(pImage->cbTable % pImage->cbCluster));
875	#if defined(RT_BIG_ENDIAN)
876	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
877	if (paL1TblImg)
878	{
879	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
880	pImage->cTableEntries);
881	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
882	pImage->offL1Table, paL1TblImg,
883	pImage->cbTable, pIoCtx, NULL, NULL);
884	RTMemFree(paL1TblImg);
885	}
886	else
887	rc = VERR_NO_MEMORY;
888	#else
889	/* Write L1 table directly. */
890	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
891	pImage->offL1Table, pImage->paL1Table,
892	pImage->cbTable, pIoCtx, NULL, NULL);
893	#endif
894	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
895	{
896	/* Write header. */
897	qedHdrConvertFromHostEndianess(pImage, &Header);
898	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
899	0, &Header, sizeof(Header),
900	pIoCtx, NULL, NULL);
901	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
902	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
903	pIoCtx, NULL, NULL);
904	}
905	}
906
907	return rc;
908	}
909
910	/**
911	* Checks whether the given cluster offset is valid.
912	*
913	* @returns Whether the given cluster offset is valid.
914	* @param offCluster The table offset to check.
915	* @param cbFile The real file size of the image.
916	* @param cbCluster The cluster size in bytes.
917	*/
918	DECLINLINE(bool) qedIsClusterOffsetValid(uint64_t offCluster, uint64_t cbFile, size_t cbCluster)
919	{
920	return (offCluster <= cbFile - cbCluster)
921	&& !(offCluster & (cbCluster - 1));
922	}
923
924	/**
925	* Checks whether the given table offset is valid.
926	*
927	* @returns Whether the given table offset is valid.
928	* @param offTbl The table offset to check.
929	* @param cbFile The real file size of the image.
930	* @param cbTable The table size in bytes.
931	* @param cbCluster The cluster size in bytes.
932	*/
933	DECLINLINE(bool) qedIsTblOffsetValid(uint64_t offTbl, uint64_t cbFile, size_t cbTable, size_t cbCluster)
934	{
935	return (offTbl <= cbFile - cbTable)
936	&& !(offTbl & (cbCluster - 1));
937	}
938
939	/**
940	* Sets the specified range in the cluster bitmap checking whether any of the clusters is already
941	* used before.
942	*
943	* @returns Whether the range was clear and is set now.
944	* @param pvClusterBitmap The cluster bitmap to use.
945	* @param offClusterStart The first cluster to check and set.
946	* @param offClusterEnd The first cluster to not check and set anymore.
947	*/
948	static bool qedClusterBitmapCheckAndSet(void *pvClusterBitmap, uint32_t offClusterStart, uint32_t offClusterEnd)
949	{
950	for (uint32_t offCluster = offClusterStart; offCluster < offClusterEnd; offCluster++)
951	if (ASMBitTest(pvClusterBitmap, offCluster))
952	return false;
953
954	ASMBitSetRange(pvClusterBitmap, offClusterStart, offClusterEnd);
955	return true;
956	}
957
958	/**
959	* Checks the given image for consistency, usually called when the
960	* QED_FEATURE_NEED_CHECK bit is set.
961	*
962	* @returns VBox status code.
963	* @retval VINF_SUCCESS when the image can be accessed.
964	* @param pImage The image instance data.
965	* @param pHeader The header to use for checking.
966	*
967	* @note It is not required that the image state is fully initialized Only
968	* The I/O interface and storage handle need to be valid.
969	* @note The header must be converted to the host CPU endian format already
970	* and should be validated already.
971	*/
972	static int qedCheckImage(PQEDIMAGE pImage, PQedHeader pHeader)
973	{
974	uint64_t cbFile;
975	uint32_t cbTable;
976	uint32_t cTableEntries;
977	uint64_t *paL1Tbl = NULL;
978	uint64_t *paL2Tbl = NULL;
979	void *pvClusterBitmap = NULL;
980	uint32_t offClusterStart;
981	int rc = VINF_SUCCESS;
982
983	pImage->cbCluster = pHeader->u32ClusterSize;
984	cbTable = pHeader->u32TableSize * pHeader->u32ClusterSize;
985	cTableEntries = cbTable / sizeof(uint64_t);
986
987	do
988	{
989	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
990	if (RT_FAILURE(rc))
991	{
992	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
993	N_("Qed: Querying the file size of image '%s' failed"),
994	pImage->pszFilename);
995	break;
996	}
997
998	/* Allocate L1 table. */
999	paL1Tbl = (uint64_t *)RTMemAllocZ(cbTable);
1000	if (!paL1Tbl)
1001	{
1002	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1003	N_("Qed: Allocating memory for the L1 table for image '%s' failed"),
1004	pImage->pszFilename);
1005	break;
1006	}
1007
1008	paL2Tbl = (uint64_t *)RTMemAllocZ(cbTable);
1009	if (!paL2Tbl)
1010	{
1011	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1012	N_("Qed: Allocating memory for the L2 table for image '%s' failed"),
1013	pImage->pszFilename);
1014	break;
1015	}
1016
1017	pvClusterBitmap = RTMemAllocZ(cbFile / pHeader->u32ClusterSize / 8);
1018	if (!pvClusterBitmap)
1019	{
1020	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1021	N_("Qed: Allocating memory for the cluster bitmap for image '%s' failed"),
1022	pImage->pszFilename);
1023	break;
1024	}
1025
1026	/* Validate L1 table offset. */
1027	if (!qedIsTblOffsetValid(pHeader->u64OffL1Table, cbFile, cbTable, pHeader->u32ClusterSize))
1028	{
1029	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1030	N_("Qed: L1 table offset of image '%s' is corrupt (%llu)"),
1031	pImage->pszFilename, pHeader->u64OffL1Table);
1032	break;
1033	}
1034
1035	/* Read L1 table. */
1036	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1037	pHeader->u64OffL1Table, paL1Tbl, cbTable);
1038	if (RT_FAILURE(rc))
1039	{
1040	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1041	N_("Qed: Reading the L1 table from image '%s' failed"),
1042	pImage->pszFilename);
1043	break;
1044	}
1045
1046	/* Mark the L1 table in cluster bitmap. */
1047	ASMBitSet(pvClusterBitmap, 0); /* Header is always in cluster 0. */
1048	offClusterStart = qedByte2Cluster(pImage, pHeader->u64OffL1Table);
1049	bool fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1050	Assert(fSet);
1051
1052	/* Scan the L1 and L2 tables for invalid entries. */
1053	qedTableConvertToHostEndianess(paL1Tbl, cTableEntries);
1054
1055	for (unsigned iL1 = 0; iL1 < cTableEntries; iL1++)
1056	{
1057	if (!paL1Tbl[iL1])
1058	continue; /* Skip unallocated clusters. */
1059
1060	if (!qedIsTblOffsetValid(paL1Tbl[iL1], cbFile, cbTable, pHeader->u32ClusterSize))
1061	{
1062	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1063	N_("Qed: Entry %d of the L1 table from image '%s' is invalid (%llu)"),
1064	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1065	break;
1066	}
1067
1068	/* Now check that the clusters are not allocated already. */
1069	offClusterStart = qedByte2Cluster(pImage, paL1Tbl[iL1]);
1070	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1071	if (!fSet)
1072	{
1073	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1074	N_("Qed: Entry %d of the L1 table from image '%s' points to a already used cluster (%llu)"),
1075	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1076	break;
1077	}
1078
1079	/* Read the linked L2 table and check it. */
1080	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1081	paL1Tbl[iL1], paL2Tbl, cbTable);
1082	if (RT_FAILURE(rc))
1083	{
1084	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1085	N_("Qed: Reading the L2 table from image '%s' failed"),
1086	pImage->pszFilename);
1087	break;
1088	}
1089
1090	/* Check all L2 entries. */
1091	for (unsigned iL2 = 0; iL2 < cTableEntries; iL2++)
1092	{
1093	if (paL2Tbl[iL2])
1094	continue; /* Skip unallocated clusters. */
1095
1096	if (!qedIsClusterOffsetValid(paL2Tbl[iL2], cbFile, pHeader->u32ClusterSize))
1097	{
1098	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1099	N_("Qed: Entry %d of the L2 table from image '%s' is invalid (%llu)"),
1100	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1101	break;
1102	}
1103
1104	/* Now check that the clusters are not allocated already. */
1105	offClusterStart = qedByte2Cluster(pImage, paL2Tbl[iL2]);
1106	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + 1);
1107	if (!fSet)
1108	{
1109	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1110	N_("Qed: Entry %d of the L2 table from image '%s' points to a already used cluster (%llu)"),
1111	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1112	break;
1113	}
1114	}
1115	}
1116	} while(0);
1117
1118	if (paL1Tbl)
1119	RTMemFree(paL1Tbl);
1120	if (paL2Tbl)
1121	RTMemFree(paL2Tbl);
1122	if (pvClusterBitmap)
1123	RTMemFree(pvClusterBitmap);
1124
1125	return rc;
1126	}
1127
1128	/**
1129	* Internal. Free all allocated space for representing an image except pImage,
1130	* and optionally delete the image from disk.
1131	*/
1132	static int qedFreeImage(PQEDIMAGE pImage, bool fDelete)
1133	{
1134	int rc = VINF_SUCCESS;
1135
1136	/* Freeing a never allocated image (e.g. because the open failed) is
1137	* not signalled as an error. After all nothing bad happens. */
1138	if (pImage)
1139	{
1140	if (pImage->pStorage)
1141	{
1142	/* No point updating the file that is deleted anyway. */
1143	if (!fDelete)
1144	qedFlushImage(pImage);
1145
1146	rc = vdIfIoIntFileClose(pImage->pIfIo, pImage->pStorage);
1147	pImage->pStorage = NULL;
1148	}
1149
1150	if (pImage->paL1Table)
1151	RTMemFree(pImage->paL1Table);
1152
1153	if (pImage->pszBackingFilename)
1154	{
1155	RTMemFree(pImage->pszBackingFilename);
1156	pImage->pszBackingFilename = NULL;
1157	}
1158
1159	qedL2TblCacheDestroy(pImage);
1160
1161	if (fDelete && pImage->pszFilename)
1162	vdIfIoIntFileDelete(pImage->pIfIo, pImage->pszFilename);
1163	}
1164
1165	LogFlowFunc(("returns %Rrc\n", rc));
1166	return rc;
1167	}
1168
1169	/**
1170	* Internal: Open an image, constructing all necessary data structures.
1171	*/
1172	static int qedOpenImage(PQEDIMAGE pImage, unsigned uOpenFlags)
1173	{
1174	int rc;
1175
1176	pImage->uOpenFlags = uOpenFlags;
1177
1178	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1179	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1180	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1181
1182	/*
1183	* Create the L2 cache before opening the image so we can call qedFreeImage()
1184	* even if opening the image file fails.
1185	*/
1186	rc = qedL2TblCacheCreate(pImage);
1187	if (RT_FAILURE(rc))
1188	{
1189	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1190	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1191	pImage->pszFilename);
1192
1193	goto out;
1194	}
1195
1196	/*
1197	* Open the image.
1198	*/
1199	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename,
1200	VDOpenFlagsToFileOpenFlags(uOpenFlags,
1201	false /* fCreate */),
1202	&pImage->pStorage);
1203	if (RT_FAILURE(rc))
1204	{
1205	/* Do NOT signal an appropriate error here, as the VD layer has the
1206	* choice of retrying the open if it failed. */
1207	goto out;
1208	}
1209
1210	uint64_t cbFile;
1211	QedHeader Header;
1212	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1213	if (RT_FAILURE(rc))
1214	goto out;
1215	if (cbFile > sizeof(Header))
1216	{
1217	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, 0, &Header, sizeof(Header));
1218	if ( RT_SUCCESS(rc)
1219	&& qedHdrConvertToHostEndianess(&Header))
1220	{
1221	if ( !(Header.u64FeatureFlags & ~QED_FEATURE_MASK)
1222	&& !(Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE_NO_PROBE))
1223	{
1224	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1225	{
1226	/* Image needs checking. */
1227	if (!(uOpenFlags & VD_OPEN_FLAGS_READONLY))
1228	rc = qedCheckImage(pImage, &Header);
1229	else
1230	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1231	N_("Qed: Image '%s' needs checking but is opened readonly"),
1232	pImage->pszFilename);
1233	}
1234
1235	if ( RT_SUCCESS(rc)
1236	&& (Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE))
1237	{
1238	/* Load backing filename from image. */
1239	pImage->pszBackingFilename = (char )RTMemAllocZ(Header.u32BackingFilenameSize + 1); / +1 for \0 terminator. */
1240	if (pImage->pszBackingFilename)
1241	{
1242	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1243	pImage->offBackingFilename = Header.u32OffBackingFilename;
1244	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1245	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1246	Header.u32BackingFilenameSize);
1247	}
1248	else
1249	rc = VERR_NO_MEMORY;
1250	}
1251
1252	if (RT_SUCCESS(rc))
1253	{
1254	pImage->cbImage = cbFile;
1255	pImage->cbCluster = Header.u32ClusterSize;
1256	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1257	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1258	pImage->offL1Table = Header.u64OffL1Table;
1259	pImage->cbSize = Header.u64Size;
1260	qedTableMasksInit(pImage);
1261
1262	/* Allocate L1 table. */
1263	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1264	if (pImage->paL1Table)
1265	{
1266	/* Read from the image. */
1267	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1268	pImage->offL1Table, pImage->paL1Table,
1269	pImage->cbTable);
1270	if (RT_SUCCESS(rc))
1271	{
1272	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1273
1274	/* If the consistency check succeeded, clear the flag by flushing the image. */
1275	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1276	rc = qedFlushImage(pImage);
1277	}
1278	else
1279	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1280	N_("Qed: Reading the L1 table for image '%s' failed"),
1281	pImage->pszFilename);
1282	}
1283	else
1284	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1285	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1286	pImage->pszFilename);
1287	}
1288	}
1289	else
1290	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1291	N_("Qed: The image '%s' makes use of unsupported features"),
1292	pImage->pszFilename);
1293	}
1294	else if (RT_SUCCESS(rc))
1295	rc = VERR_VD_GEN_INVALID_HEADER;
1296	}
1297	else
1298	rc = VERR_VD_GEN_INVALID_HEADER;
1299
1300	out:
1301	if (RT_FAILURE(rc))
1302	qedFreeImage(pImage, false);
1303	return rc;
1304	}
1305
1306	/**
1307	* Internal: Create a qed image.
1308	*/
1309	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1310	unsigned uImageFlags, const char *pszComment,
1311	PCVDGEOMETRY pPCHSGeometry,
1312	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1313	PFNVDPROGRESS pfnProgress, void *pvUser,
1314	unsigned uPercentStart, unsigned uPercentSpan)
1315	{
1316	int rc;
1317	int32_t fOpen;
1318
1319	if (uImageFlags & VD_IMAGE_FLAGS_FIXED)
1320	{
1321	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1322	goto out;
1323	}
1324
1325	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1326	pImage->uImageFlags = uImageFlags;
1327	pImage->PCHSGeometry = *pPCHSGeometry;
1328	pImage->LCHSGeometry = *pLCHSGeometry;
1329
1330	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1331	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1332	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1333
1334	/* Create image file. */
1335	fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1336	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1337	if (RT_FAILURE(rc))
1338	{
1339	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1340	goto out;
1341	}
1342
1343	/* Init image state. */
1344	pImage->cbSize = cbSize;
1345	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1346	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1347	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1348	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1349	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1350	pImage->cbBackingFilename = 0;
1351	pImage->offBackingFilename = 0;
1352	qedTableMasksInit(pImage);
1353
1354	/* Init L1 table. */
1355	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1356	if (!pImage->paL1Table)
1357	{
1358	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1359	pImage->pszFilename);
1360	goto out;
1361	}
1362
1363	rc = qedL2TblCacheCreate(pImage);
1364	if (RT_FAILURE(rc))
1365	{
1366	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1367	pImage->pszFilename);
1368	goto out;
1369	}
1370
1371	if (RT_SUCCESS(rc) && pfnProgress)
1372	pfnProgress(pvUser, uPercentStart + uPercentSpan * 98 / 100);
1373
1374	rc = qedFlushImage(pImage);
1375
1376	out:
1377	if (RT_SUCCESS(rc) && pfnProgress)
1378	pfnProgress(pvUser, uPercentStart + uPercentSpan);
1379
1380	if (RT_FAILURE(rc))
1381	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1382	return rc;
1383	}
1384
1385	/**
1386	* Rollback anything done during async cluster allocation.
1387	*
1388	* @returns VBox status code.
1389	* @param pImage The image instance data.
1390	* @param pIoCtx The I/O context.
1391	* @param pClusterAlloc The cluster allocation to rollback.
1392	*/
1393	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1394	{
1395	int rc = VINF_SUCCESS;
1396
1397	switch (pClusterAlloc->enmAllocState)
1398	{
1399	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1400	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1401	{
1402	/* Assumption right now is that the L1 table is not modified if the link fails. */
1403	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1404	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1405	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1406	break;
1407	}
1408	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1409	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1410	{
1411	/* Assumption right now is that the L2 table is not modified if the link fails. */
1412	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1413	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1414	break;
1415	}
1416	default:
1417	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1418	rc = VERR_INVALID_STATE;
1419	}
1420
1421	RTMemFree(pClusterAlloc);
1422	return rc;
1423	}
1424
1425	/**
1426	* Updates the state of the async cluster allocation.
1427	*
1428	* @returns VBox status code.
1429	* @param pBackendData The opaque backend data.
1430	* @param pIoCtx I/O context associated with this request.
1431	* @param pvUser Opaque user data passed during a read/write request.
1432	* @param rcReq Status code for the completed request.
1433	*/
1434	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1435	{
1436	int rc = VINF_SUCCESS;
1437	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1438	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1439
1440	if (RT_FAILURE(rcReq))
1441	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1442
1443	AssertPtr(pClusterAlloc->pL2Entry);
1444
1445	switch (pClusterAlloc->enmAllocState)
1446	{
1447	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1448	{
1449	uint64_t offUpdateLe = RT_H2LE_U64(pClusterAlloc->pL2Entry->offL2Tbl);
1450
1451	/* Update the link in the on disk L1 table now. */
1452	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1453	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1454	pImage->offL1Table + pClusterAlloc->idxL1*sizeof(uint64_t),
1455	&offUpdateLe, sizeof(uint64_t), pIoCtx,
1456	qedAsyncClusterAllocUpdate, pClusterAlloc);
1457	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1458	break;
1459	else if (RT_FAILURE(rc))
1460	{
1461	/* Rollback. */
1462	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1463	break;
1464	}
1465	/* Success, fall through. */
1466	}
1467	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1468	{
1469	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1470	uint64_t offData = qedClusterAllocate(pImage, 1);
1471
1472	/* Update the link in the in memory L1 table now. */
1473	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1474	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1475
1476	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1477	pClusterAlloc->cbImageOld = offData;
1478	pClusterAlloc->offClusterNew = offData;
1479
1480	/* Write data. */
1481	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1482	offData, pIoCtx, pClusterAlloc->cbToWrite,
1483	qedAsyncClusterAllocUpdate, pClusterAlloc);
1484	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1485	break;
1486	else if (RT_FAILURE(rc))
1487	{
1488	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1489	RTMemFree(pClusterAlloc);
1490	break;
1491	}
1492	}
1493	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1494	{
1495	uint64_t offUpdateLe = RT_H2LE_U64(pClusterAlloc->offClusterNew);
1496
1497	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1498
1499	/* Link L2 table and update it. */
1500	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1501	pImage->paL1Table[pClusterAlloc->idxL1] + pClusterAlloc->idxL2*sizeof(uint64_t),
1502	&offUpdateLe, sizeof(uint64_t), pIoCtx,
1503	qedAsyncClusterAllocUpdate, pClusterAlloc);
1504	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1505	break;
1506	else if (RT_FAILURE(rc))
1507	{
1508	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1509	RTMemFree(pClusterAlloc);
1510	break;
1511	}
1512	}
1513	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1514	{
1515	/* Everything done without errors, signal completion. */
1516	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1517	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1518	RTMemFree(pClusterAlloc);
1519	rc = VINF_SUCCESS;
1520	break;
1521	}
1522	default:
1523	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1524	pClusterAlloc->enmAllocState));
1525	}
1526
1527	return rc;
1528	}
1529
1530	/** @copydoc VBOXHDDBACKEND::pfnCheckIfValid */
1531	static int qedCheckIfValid(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1532	PVDINTERFACE pVDIfsImage, VDTYPE *penmType)
1533	{
1534	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1535	PVDIOSTORAGE pStorage = NULL;
1536	uint64_t cbFile;
1537	int rc = VINF_SUCCESS;
1538
1539	/* Get I/O interface. */
1540	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1541	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1542
1543	if ( !VALID_PTR(pszFilename)
1544	\|\| !*pszFilename)
1545	{
1546	rc = VERR_INVALID_PARAMETER;
1547	goto out;
1548	}
1549
1550	/*
1551	* Open the file and read the footer.
1552	*/
1553	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1554	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1555	false /* fCreate */),
1556	&pStorage);
1557	if (RT_SUCCESS(rc))
1558	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1559
1560	if ( RT_SUCCESS(rc)
1561	&& cbFile > sizeof(QedHeader))
1562	{
1563	QedHeader Header;
1564
1565	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1566	if ( RT_SUCCESS(rc)
1567	&& qedHdrConvertToHostEndianess(&Header))
1568	{
1569	*penmType = VDTYPE_HDD;
1570	rc = VINF_SUCCESS;
1571	}
1572	else
1573	rc = VERR_VD_GEN_INVALID_HEADER;
1574	}
1575	else
1576	rc = VERR_VD_GEN_INVALID_HEADER;
1577
1578	if (pStorage)
1579	vdIfIoIntFileClose(pIfIo, pStorage);
1580
1581	out:
1582	LogFlowFunc(("returns %Rrc\n", rc));
1583	return rc;
1584	}
1585
1586	/** @copydoc VBOXHDDBACKEND::pfnOpen */
1587	static int qedOpen(const char *pszFilename, unsigned uOpenFlags,
1588	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1589	VDTYPE enmType, void **ppBackendData)
1590	{
1591	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n", pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1592	int rc;
1593	PQEDIMAGE pImage;
1594
1595	NOREF(enmType); /*< @todo r=klaus make use of the type info. /
1596
1597	/* Check open flags. All valid flags are supported. */
1598	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1599	{
1600	rc = VERR_INVALID_PARAMETER;
1601	goto out;
1602	}
1603
1604	/* Check remaining arguments. */
1605	if ( !VALID_PTR(pszFilename)
1606	\|\| !*pszFilename)
1607	{
1608	rc = VERR_INVALID_PARAMETER;
1609	goto out;
1610	}
1611
1612
1613	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1614	if (!pImage)
1615	{
1616	rc = VERR_NO_MEMORY;
1617	goto out;
1618	}
1619	pImage->pszFilename = pszFilename;
1620	pImage->pStorage = NULL;
1621	pImage->pVDIfsDisk = pVDIfsDisk;
1622	pImage->pVDIfsImage = pVDIfsImage;
1623
1624	rc = qedOpenImage(pImage, uOpenFlags);
1625	if (RT_SUCCESS(rc))
1626	*ppBackendData = pImage;
1627	else
1628	RTMemFree(pImage);
1629
1630	out:
1631	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1632	return rc;
1633	}
1634
1635	/** @copydoc VBOXHDDBACKEND::pfnCreate */
1636	static int qedCreate(const char *pszFilename, uint64_t cbSize,
1637	unsigned uImageFlags, const char *pszComment,
1638	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1639	PCRTUUID pUuid, unsigned uOpenFlags,
1640	unsigned uPercentStart, unsigned uPercentSpan,
1641	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1642	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1643	void **ppBackendData)
1644	{
1645	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",
1646	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1647	int rc;
1648	PQEDIMAGE pImage;
1649
1650	PFNVDPROGRESS pfnProgress = NULL;
1651	void *pvUser = NULL;
1652	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1653	if (pIfProgress)
1654	{
1655	pfnProgress = pIfProgress->pfnProgress;
1656	pvUser = pIfProgress->Core.pvUser;
1657	}
1658
1659	/* Check the VD container type. */
1660	if (enmType != VDTYPE_HDD)
1661	{
1662	rc = VERR_VD_INVALID_TYPE;
1663	goto out;
1664	}
1665
1666	/* Check open flags. All valid flags are supported. */
1667	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1668	{
1669	rc = VERR_INVALID_PARAMETER;
1670	goto out;
1671	}
1672
1673	/* Check remaining arguments. */
1674	if ( !VALID_PTR(pszFilename)
1675	\|\| !*pszFilename
1676	\|\| !VALID_PTR(pPCHSGeometry)
1677	\|\| !VALID_PTR(pLCHSGeometry))
1678	{
1679	rc = VERR_INVALID_PARAMETER;
1680	goto out;
1681	}
1682
1683	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1684	if (!pImage)
1685	{
1686	rc = VERR_NO_MEMORY;
1687	goto out;
1688	}
1689	pImage->pszFilename = pszFilename;
1690	pImage->pStorage = NULL;
1691	pImage->pVDIfsDisk = pVDIfsDisk;
1692	pImage->pVDIfsImage = pVDIfsImage;
1693
1694	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1695	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1696	pfnProgress, pvUser, uPercentStart, uPercentSpan);
1697	if (RT_SUCCESS(rc))
1698	{
1699	/* So far the image is opened in read/write mode. Make sure the
1700	* image is opened in read-only mode if the caller requested that. */
1701	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1702	{
1703	qedFreeImage(pImage, false);
1704	rc = qedOpenImage(pImage, uOpenFlags);
1705	if (RT_FAILURE(rc))
1706	{
1707	RTMemFree(pImage);
1708	goto out;
1709	}
1710	}
1711	*ppBackendData = pImage;
1712	}
1713	else
1714	RTMemFree(pImage);
1715
1716	out:
1717	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1718	return rc;
1719	}
1720
1721	/** @copydoc VBOXHDDBACKEND::pfnRename */
1722	static int qedRename(void pBackendData, const char pszFilename)
1723	{
1724	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1725	int rc = VINF_SUCCESS;
1726	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1727
1728	/* Check arguments. */
1729	if ( !pImage
1730	\|\| !pszFilename
1731	\|\| !*pszFilename)
1732	{
1733	rc = VERR_INVALID_PARAMETER;
1734	goto out;
1735	}
1736
1737	/* Close the image. */
1738	rc = qedFreeImage(pImage, false);
1739	if (RT_FAILURE(rc))
1740	goto out;
1741
1742	/* Rename the file. */
1743	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1744	if (RT_FAILURE(rc))
1745	{
1746	/* The move failed, try to reopen the original image. */
1747	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1748	if (RT_FAILURE(rc2))
1749	rc = rc2;
1750
1751	goto out;
1752	}
1753
1754	/* Update pImage with the new information. */
1755	pImage->pszFilename = pszFilename;
1756
1757	/* Open the old image with new name. */
1758	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1759	if (RT_FAILURE(rc))
1760	goto out;
1761
1762	out:
1763	LogFlowFunc(("returns %Rrc\n", rc));
1764	return rc;
1765	}
1766
1767	/** @copydoc VBOXHDDBACKEND::pfnClose */
1768	static int qedClose(void *pBackendData, bool fDelete)
1769	{
1770	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1771	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1772	int rc;
1773
1774	rc = qedFreeImage(pImage, fDelete);
1775	RTMemFree(pImage);
1776
1777	LogFlowFunc(("returns %Rrc\n", rc));
1778	return rc;
1779	}
1780
1781	static int qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1782	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1783	{
1784	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1785	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1786	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1787	uint32_t offCluster = 0;
1788	uint32_t idxL1 = 0;
1789	uint32_t idxL2 = 0;
1790	uint64_t offFile = 0;
1791	int rc;
1792
1793	AssertPtr(pImage);
1794	Assert(uOffset % 512 == 0);
1795	Assert(cbToRead % 512 == 0);
1796
1797	if (!VALID_PTR(pIoCtx) \|\| !cbToRead)
1798	{
1799	rc = VERR_INVALID_PARAMETER;
1800	goto out;
1801	}
1802
1803	if ( uOffset + cbToRead > pImage->cbSize
1804	\|\| cbToRead == 0)
1805	{
1806	rc = VERR_INVALID_PARAMETER;
1807	goto out;
1808	}
1809
1810	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1811
1812	/* Clip read size to remain in the cluster. */
1813	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1814
1815	/* Get offset in image. */
1816	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1817	if (RT_SUCCESS(rc))
1818	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1819	pIoCtx, cbToRead);
1820
1821	if ( ( RT_SUCCESS(rc)
1822	\|\| rc == VERR_VD_BLOCK_FREE
1823	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1824	&& pcbActuallyRead)
1825	*pcbActuallyRead = cbToRead;
1826
1827	out:
1828	LogFlowFunc(("returns %Rrc\n", rc));
1829	return rc;
1830	}
1831
1832	static int qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1833	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1834	size_t *pcbPostRead, unsigned fWrite)
1835	{
1836	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1837	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1838	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1839	uint32_t offCluster = 0;
1840	uint32_t idxL1 = 0;
1841	uint32_t idxL2 = 0;
1842	uint64_t offImage = 0;
1843	int rc = VINF_SUCCESS;
1844
1845	AssertPtr(pImage);
1846	Assert(!(uOffset % 512));
1847	Assert(!(cbToWrite % 512));
1848
1849	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
1850	{
1851	rc = VERR_VD_IMAGE_READ_ONLY;
1852	goto out;
1853	}
1854
1855	if (!VALID_PTR(pIoCtx) \|\| !cbToWrite)
1856	{
1857	rc = VERR_INVALID_PARAMETER;
1858	goto out;
1859	}
1860
1861	if ( uOffset + cbToWrite > pImage->cbSize
1862	\|\| cbToWrite == 0)
1863	{
1864	rc = VERR_INVALID_PARAMETER;
1865	goto out;
1866	}
1867
1868	/* Convert offset to L1, L2 index and cluster offset. */
1869	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1870
1871	/* Clip write size to remain in the cluster. */
1872	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1873	Assert(!(cbToWrite % 512));
1874
1875	/* Get offset in image. */
1876	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1877	if (RT_SUCCESS(rc))
1878	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1879	offImage, pIoCtx, cbToWrite, NULL, NULL);
1880	else if (rc == VERR_VD_BLOCK_FREE)
1881	{
1882	if ( cbToWrite == pImage->cbCluster
1883	&& !(fWrite & VD_WRITE_NO_ALLOC))
1884	{
1885	PQEDL2CACHEENTRY pL2Entry = NULL;
1886
1887	/* Full cluster write to previously unallocated cluster.
1888	* Allocate cluster and write data. */
1889	Assert(!offCluster);
1890
1891	do
1892	{
1893	uint64_t idxUpdateLe = 0;
1894
1895	/* Check if we have to allocate a new cluster for L2 tables. */
1896	if (!pImage->paL1Table[idxL1])
1897	{
1898	uint64_t offL2Tbl;
1899	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1900
1901	/* Allocate new async cluster allocation state. */
1902	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1903	if (RT_UNLIKELY(!pL2ClusterAlloc))
1904	{
1905	rc = VERR_NO_MEMORY;
1906	break;
1907	}
1908
1909	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1910	if (!pL2Entry)
1911	{
1912	rc = VERR_NO_MEMORY;
1913	RTMemFree(pL2ClusterAlloc);
1914	break;
1915	}
1916
1917	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1918	pL2Entry->offL2Tbl = offL2Tbl;
1919	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1920
1921	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1922	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1923	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1924	pL2ClusterAlloc->idxL1 = idxL1;
1925	pL2ClusterAlloc->idxL2 = idxL2;
1926	pL2ClusterAlloc->cbToWrite = cbToWrite;
1927	pL2ClusterAlloc->pL2Entry = pL2Entry;
1928
1929	/*
1930	* Write the L2 table first and link to the L1 table afterwards.
1931	* If something unexpected happens the worst case which can happen
1932	* is a leak of some clusters.
1933	*/
1934	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1935	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1936	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1937	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1938	break;
1939	else if (RT_FAILURE(rc))
1940	{
1941	RTMemFree(pL2ClusterAlloc);
1942	qedL2TblCacheEntryFree(pImage, pL2Entry);
1943	break;
1944	}
1945
1946	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1947	}
1948	else
1949	{
1950	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1951	&pL2Entry);
1952
1953	if (RT_SUCCESS(rc))
1954	{
1955	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1956
1957	/* Allocate new async cluster allocation state. */
1958	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1959	if (RT_UNLIKELY(!pDataClusterAlloc))
1960	{
1961	rc = VERR_NO_MEMORY;
1962	break;
1963	}
1964
1965	/* Allocate new cluster for the data. */
1966	uint64_t offData = qedClusterAllocate(pImage, 1);
1967
1968	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1969	pDataClusterAlloc->cbImageOld = offData;
1970	pDataClusterAlloc->offClusterNew = offData;
1971	pDataClusterAlloc->idxL1 = idxL1;
1972	pDataClusterAlloc->idxL2 = idxL2;
1973	pDataClusterAlloc->cbToWrite = cbToWrite;
1974	pDataClusterAlloc->pL2Entry = pL2Entry;
1975
1976	/* Write data. */
1977	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1978	offData, pIoCtx, cbToWrite,
1979	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1980	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1981	break;
1982	else if (RT_FAILURE(rc))
1983	{
1984	RTMemFree(pDataClusterAlloc);
1985	break;
1986	}
1987
1988	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1989	}
1990	}
1991
1992	} while (0);
1993
1994	*pcbPreRead = 0;
1995	*pcbPostRead = 0;
1996	}
1997	else
1998	{
1999	/* Trying to do a partial write to an unallocated cluster. Don't do
2000	* anything except letting the upper layer know what to do. */
2001	*pcbPreRead = offCluster;
2002	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
2003	}
2004	}
2005
2006	if (pcbWriteProcess)
2007	*pcbWriteProcess = cbToWrite;
2008
2009
2010	out:
2011	LogFlowFunc(("returns %Rrc\n", rc));
2012	return rc;
2013	}
2014
2015	static int qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
2016	{
2017	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2018	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2019	int rc = VINF_SUCCESS;
2020
2021	Assert(pImage);
2022
2023	if (VALID_PTR(pIoCtx))
2024	rc = qedFlushImageAsync(pImage, pIoCtx);
2025	else
2026	rc = VERR_INVALID_PARAMETER;
2027
2028	LogFlowFunc(("returns %Rrc\n", rc));
2029	return rc;
2030	}
2031
2032	/** @copydoc VBOXHDDBACKEND::pfnGetVersion */
2033	static unsigned qedGetVersion(void *pBackendData)
2034	{
2035	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2036	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2037
2038	AssertPtr(pImage);
2039
2040	if (pImage)
2041	return 1;
2042	else
2043	return 0;
2044	}
2045
2046	/** @copydoc VBOXHDDBACKEND::pfnGetSectorSize */
2047	static uint32_t qedGetSectorSize(void *pBackendData)
2048	{
2049	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2050	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2051	uint32_t cb = 0;
2052
2053	AssertPtr(pImage);
2054
2055	if (pImage && pImage->pStorage)
2056	cb = 512;
2057
2058	LogFlowFunc(("returns %u\n", cb));
2059	return cb;
2060	}
2061
2062	/** @copydoc VBOXHDDBACKEND::pfnGetSize */
2063	static uint64_t qedGetSize(void *pBackendData)
2064	{
2065	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2066	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2067	uint64_t cb = 0;
2068
2069	AssertPtr(pImage);
2070
2071	if (pImage && pImage->pStorage)
2072	cb = pImage->cbSize;
2073
2074	LogFlowFunc(("returns %llu\n", cb));
2075	return cb;
2076	}
2077
2078	/** @copydoc VBOXHDDBACKEND::pfnGetFileSize */
2079	static uint64_t qedGetFileSize(void *pBackendData)
2080	{
2081	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2082	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2083	uint64_t cb = 0;
2084
2085	AssertPtr(pImage);
2086
2087	if (pImage)
2088	{
2089	uint64_t cbFile;
2090	if (pImage->pStorage)
2091	{
2092	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
2093	if (RT_SUCCESS(rc))
2094	cb += cbFile;
2095	}
2096	}
2097
2098	LogFlowFunc(("returns %lld\n", cb));
2099	return cb;
2100	}
2101
2102	/** @copydoc VBOXHDDBACKEND::pfnGetPCHSGeometry */
2103	static int qedGetPCHSGeometry(void *pBackendData,
2104	PVDGEOMETRY pPCHSGeometry)
2105	{
2106	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
2107	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2108	int rc;
2109
2110	AssertPtr(pImage);
2111
2112	if (pImage)
2113	{
2114	if (pImage->PCHSGeometry.cCylinders)
2115	{
2116	*pPCHSGeometry = pImage->PCHSGeometry;
2117	rc = VINF_SUCCESS;
2118	}
2119	else
2120	rc = VERR_VD_GEOMETRY_NOT_SET;
2121	}
2122	else
2123	rc = VERR_VD_NOT_OPENED;
2124
2125	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2126	return rc;
2127	}
2128
2129	/** @copydoc VBOXHDDBACKEND::pfnSetPCHSGeometry */
2130	static int qedSetPCHSGeometry(void *pBackendData,
2131	PCVDGEOMETRY pPCHSGeometry)
2132	{
2133	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n", pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2134	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2135	int rc;
2136
2137	AssertPtr(pImage);
2138
2139	if (pImage)
2140	{
2141	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2142	{
2143	rc = VERR_VD_IMAGE_READ_ONLY;
2144	goto out;
2145	}
2146
2147	pImage->PCHSGeometry = *pPCHSGeometry;
2148	rc = VINF_SUCCESS;
2149	}
2150	else
2151	rc = VERR_VD_NOT_OPENED;
2152
2153	out:
2154	LogFlowFunc(("returns %Rrc\n", rc));
2155	return rc;
2156	}
2157
2158	/** @copydoc VBOXHDDBACKEND::pfnGetLCHSGeometry */
2159	static int qedGetLCHSGeometry(void *pBackendData,
2160	PVDGEOMETRY pLCHSGeometry)
2161	{
2162	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2163	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2164	int rc;
2165
2166	AssertPtr(pImage);
2167
2168	if (pImage)
2169	{
2170	if (pImage->LCHSGeometry.cCylinders)
2171	{
2172	*pLCHSGeometry = pImage->LCHSGeometry;
2173	rc = VINF_SUCCESS;
2174	}
2175	else
2176	rc = VERR_VD_GEOMETRY_NOT_SET;
2177	}
2178	else
2179	rc = VERR_VD_NOT_OPENED;
2180
2181	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2182	return rc;
2183	}
2184
2185	/** @copydoc VBOXHDDBACKEND::pfnSetLCHSGeometry */
2186	static int qedSetLCHSGeometry(void *pBackendData,
2187	PCVDGEOMETRY pLCHSGeometry)
2188	{
2189	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData, pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2190	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2191	int rc;
2192
2193	AssertPtr(pImage);
2194
2195	if (pImage)
2196	{
2197	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2198	{
2199	rc = VERR_VD_IMAGE_READ_ONLY;
2200	goto out;
2201	}
2202
2203	pImage->LCHSGeometry = *pLCHSGeometry;
2204	rc = VINF_SUCCESS;
2205	}
2206	else
2207	rc = VERR_VD_NOT_OPENED;
2208
2209	out:
2210	LogFlowFunc(("returns %Rrc\n", rc));
2211	return rc;
2212	}
2213
2214	/** @copydoc VBOXHDDBACKEND::pfnGetImageFlags */
2215	static unsigned qedGetImageFlags(void *pBackendData)
2216	{
2217	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2218	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2219	unsigned uImageFlags;
2220
2221	AssertPtr(pImage);
2222
2223	if (pImage)
2224	uImageFlags = pImage->uImageFlags;
2225	else
2226	uImageFlags = 0;
2227
2228	LogFlowFunc(("returns %#x\n", uImageFlags));
2229	return uImageFlags;
2230	}
2231
2232	/** @copydoc VBOXHDDBACKEND::pfnGetOpenFlags */
2233	static unsigned qedGetOpenFlags(void *pBackendData)
2234	{
2235	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2236	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2237	unsigned uOpenFlags;
2238
2239	AssertPtr(pImage);
2240
2241	if (pImage)
2242	uOpenFlags = pImage->uOpenFlags;
2243	else
2244	uOpenFlags = 0;
2245
2246	LogFlowFunc(("returns %#x\n", uOpenFlags));
2247	return uOpenFlags;
2248	}
2249
2250	/** @copydoc VBOXHDDBACKEND::pfnSetOpenFlags */
2251	static int qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2252	{
2253	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2254	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2255	int rc;
2256
2257	/* Image must be opened and the new flags must be valid. */
2258	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2259	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2260	{
2261	rc = VERR_INVALID_PARAMETER;
2262	goto out;
2263	}
2264
2265	/* Implement this operation via reopening the image. */
2266	rc = qedFreeImage(pImage, false);
2267	if (RT_FAILURE(rc))
2268	goto out;
2269	rc = qedOpenImage(pImage, uOpenFlags);
2270
2271	out:
2272	LogFlowFunc(("returns %Rrc\n", rc));
2273	return rc;
2274	}
2275
2276	/** @copydoc VBOXHDDBACKEND::pfnGetComment */
2277	static int qedGetComment(void pBackendData, char pszComment,
2278	size_t cbComment)
2279	{
2280	LogFlowFunc(("pBackendData=%#p pszComment=%#p cbComment=%zu\n", pBackendData, pszComment, cbComment));
2281	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2282	int rc;
2283
2284	AssertPtr(pImage);
2285
2286	if (pImage)
2287	rc = VERR_NOT_SUPPORTED;
2288	else
2289	rc = VERR_VD_NOT_OPENED;
2290
2291	LogFlowFunc(("returns %Rrc comment='%s'\n", rc, pszComment));
2292	return rc;
2293	}
2294
2295	/** @copydoc VBOXHDDBACKEND::pfnSetComment */
2296	static int qedSetComment(void pBackendData, const char pszComment)
2297	{
2298	LogFlowFunc(("pBackendData=%#p pszComment=\"%s\"\n", pBackendData, pszComment));
2299	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2300	int rc;
2301
2302	AssertPtr(pImage);
2303
2304	if (pImage)
2305	{
2306	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2307	rc = VERR_VD_IMAGE_READ_ONLY;
2308	else
2309	rc = VERR_NOT_SUPPORTED;
2310	}
2311	else
2312	rc = VERR_VD_NOT_OPENED;
2313
2314	LogFlowFunc(("returns %Rrc\n", rc));
2315	return rc;
2316	}
2317
2318	/** @copydoc VBOXHDDBACKEND::pfnGetUuid */
2319	static int qedGetUuid(void *pBackendData, PRTUUID pUuid)
2320	{
2321	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2322	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2323	int rc;
2324
2325	AssertPtr(pImage);
2326
2327	if (pImage)
2328	rc = VERR_NOT_SUPPORTED;
2329	else
2330	rc = VERR_VD_NOT_OPENED;
2331
2332	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2333	return rc;
2334	}
2335
2336	/** @copydoc VBOXHDDBACKEND::pfnSetUuid */
2337	static int qedSetUuid(void *pBackendData, PCRTUUID pUuid)
2338	{
2339	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2340	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2341	int rc;
2342
2343	LogFlowFunc(("%RTuuid\n", pUuid));
2344	AssertPtr(pImage);
2345
2346	if (pImage)
2347	{
2348	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2349	rc = VERR_NOT_SUPPORTED;
2350	else
2351	rc = VERR_VD_IMAGE_READ_ONLY;
2352	}
2353	else
2354	rc = VERR_VD_NOT_OPENED;
2355
2356	LogFlowFunc(("returns %Rrc\n", rc));
2357	return rc;
2358	}
2359
2360	/** @copydoc VBOXHDDBACKEND::pfnGetModificationUuid */
2361	static int qedGetModificationUuid(void *pBackendData, PRTUUID pUuid)
2362	{
2363	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2364	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2365	int rc;
2366
2367	AssertPtr(pImage);
2368
2369	if (pImage)
2370	rc = VERR_NOT_SUPPORTED;
2371	else
2372	rc = VERR_VD_NOT_OPENED;
2373
2374	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2375	return rc;
2376	}
2377
2378	/** @copydoc VBOXHDDBACKEND::pfnSetModificationUuid */
2379	static int qedSetModificationUuid(void *pBackendData, PCRTUUID pUuid)
2380	{
2381	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2382	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2383	int rc;
2384
2385	AssertPtr(pImage);
2386
2387	if (pImage)
2388	{
2389	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2390	rc = VERR_NOT_SUPPORTED;
2391	else
2392	rc = VERR_VD_IMAGE_READ_ONLY;
2393	}
2394	else
2395	rc = VERR_VD_NOT_OPENED;
2396
2397	LogFlowFunc(("returns %Rrc\n", rc));
2398	return rc;
2399	}
2400
2401	/** @copydoc VBOXHDDBACKEND::pfnGetParentUuid */
2402	static int qedGetParentUuid(void *pBackendData, PRTUUID pUuid)
2403	{
2404	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2405	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2406	int rc;
2407
2408	AssertPtr(pImage);
2409
2410	if (pImage)
2411	rc = VERR_NOT_SUPPORTED;
2412	else
2413	rc = VERR_VD_NOT_OPENED;
2414
2415	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2416	return rc;
2417	}
2418
2419	/** @copydoc VBOXHDDBACKEND::pfnSetParentUuid */
2420	static int qedSetParentUuid(void *pBackendData, PCRTUUID pUuid)
2421	{
2422	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2423	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2424	int rc;
2425
2426	AssertPtr(pImage);
2427
2428	if (pImage)
2429	{
2430	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2431	rc = VERR_NOT_SUPPORTED;
2432	else
2433	rc = VERR_VD_IMAGE_READ_ONLY;
2434	}
2435	else
2436	rc = VERR_VD_NOT_OPENED;
2437
2438	LogFlowFunc(("returns %Rrc\n", rc));
2439	return rc;
2440	}
2441
2442	/** @copydoc VBOXHDDBACKEND::pfnGetParentModificationUuid */
2443	static int qedGetParentModificationUuid(void *pBackendData, PRTUUID pUuid)
2444	{
2445	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2446	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2447	int rc;
2448
2449	AssertPtr(pImage);
2450
2451	if (pImage)
2452	rc = VERR_NOT_SUPPORTED;
2453	else
2454	rc = VERR_VD_NOT_OPENED;
2455
2456	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2457	return rc;
2458	}
2459
2460	/** @copydoc VBOXHDDBACKEND::pfnSetParentModificationUuid */
2461	static int qedSetParentModificationUuid(void *pBackendData, PCRTUUID pUuid)
2462	{
2463	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2464	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2465	int rc;
2466
2467	AssertPtr(pImage);
2468
2469	if (pImage)
2470	{
2471	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2472	rc = VERR_NOT_SUPPORTED;
2473	else
2474	rc = VERR_VD_IMAGE_READ_ONLY;
2475	}
2476	else
2477	rc = VERR_VD_NOT_OPENED;
2478
2479	LogFlowFunc(("returns %Rrc\n", rc));
2480	return rc;
2481	}
2482
2483	/** @copydoc VBOXHDDBACKEND::pfnDump */
2484	static void qedDump(void *pBackendData)
2485	{
2486	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2487
2488	AssertPtr(pImage);
2489	if (pImage)
2490	{
2491	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2492	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2493	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2494	pImage->cbSize / 512);
2495	}
2496	}
2497
2498	/** @copydoc VBOXHDDBACKEND::pfnGetParentFilename */
2499	static int qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2500	{
2501	int rc = VINF_SUCCESS;
2502	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2503
2504	AssertPtr(pImage);
2505	if (pImage)
2506	if (pImage->pszBackingFilename)
2507	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2508	else
2509	rc = VERR_NOT_SUPPORTED;
2510	else
2511	rc = VERR_VD_NOT_OPENED;
2512
2513	LogFlowFunc(("returns %Rrc\n", rc));
2514	return rc;
2515	}
2516
2517	/** @copydoc VBOXHDDBACKEND::pfnSetParentFilename */
2518	static int qedSetParentFilename(void pBackendData, const char pszParentFilename)
2519	{
2520	int rc = VINF_SUCCESS;
2521	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2522
2523	AssertPtr(pImage);
2524	if (pImage)
2525	{
2526	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2527	rc = VERR_VD_IMAGE_READ_ONLY;
2528	else if ( pImage->pszBackingFilename
2529	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2530	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2531	else
2532	{
2533	if (pImage->pszBackingFilename)
2534	RTStrFree(pImage->pszBackingFilename);
2535	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2536	if (!pImage->pszBackingFilename)
2537	rc = VERR_NO_MEMORY;
2538	else
2539	{
2540	if (!pImage->offBackingFilename)
2541	{
2542	/* Allocate new cluster. */
2543	uint64_t offData = qedClusterAllocate(pImage, 1);
2544
2545	Assert((offData & UINT32_MAX) == offData);
2546	pImage->offBackingFilename = (uint32_t)offData;
2547	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2548	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2549	offData + pImage->cbCluster);
2550	}
2551
2552	if (RT_SUCCESS(rc))
2553	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2554	pImage->offBackingFilename,
2555	pImage->pszBackingFilename,
2556	strlen(pImage->pszBackingFilename));
2557	}
2558	}
2559	}
2560	else
2561	rc = VERR_VD_NOT_OPENED;
2562
2563	LogFlowFunc(("returns %Rrc\n", rc));
2564	return rc;
2565	}
2566
2567	/** @copydoc VBOXHDDBACKEND::pfnResize */
2568	static int qedResize(void *pBackendData, uint64_t cbSize,
2569	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2570	unsigned uPercentStart, unsigned uPercentSpan,
2571	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2572	PVDINTERFACE pVDIfsOperation)
2573	{
2574	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2575	int rc = VINF_SUCCESS;
2576
2577	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
2578
2579	/* Making the image smaller is not supported at the moment. */
2580	if (cbSize < pImage->cbSize)
2581	rc = VERR_NOT_SUPPORTED;
2582	else if (cbSize > pImage->cbSize)
2583	{
2584	/*
2585	* It is enough to just update the size field in the header to complete
2586	* growing. With the default cluster and table sizes the image can be expanded
2587	* to 64TB without overflowing the L1 and L2 tables making block relocation
2588	* superfluous.
2589	* @todo: The rare case where block relocation is still required (non default
2590	* table and/or cluster size or images with more than 64TB) is not
2591	* implemented yet and resizing such an image will fail with an error.
2592	*/
2593	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2594	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2595	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2596	pImage->pszFilename);
2597	else
2598	{
2599	uint64_t cbSizeOld = pImage->cbSize;
2600
2601	pImage->cbSize = cbSize;
2602	rc = qedFlushImage(pImage);
2603	if (RT_FAILURE(rc))
2604	{
2605	pImage->cbSize = cbSizeOld; /* Restore */
2606
2607	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2608	pImage->pszFilename);
2609	}
2610	}
2611	}
2612	/* Same size doesn't change the image at all. */
2613
2614	LogFlowFunc(("returns %Rrc\n", rc));
2615	return rc;
2616	}
2617
2618
2619	const VBOXHDDBACKEND g_QedBackend =
2620	{
2621	/* pszBackendName */
2622	"QED",
2623	/* cbSize */
2624	sizeof(VBOXHDDBACKEND),
2625	/* uBackendCaps */
2626	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2627	/* paFileExtensions */
2628	s_aQedFileExtensions,
2629	/* paConfigInfo */
2630	NULL,
2631	/* pfnCheckIfValid */
2632	qedCheckIfValid,
2633	/* pfnOpen */
2634	qedOpen,
2635	/* pfnCreate */
2636	qedCreate,
2637	/* pfnRename */
2638	qedRename,
2639	/* pfnClose */
2640	qedClose,
2641	/* pfnRead */
2642	qedRead,
2643	/* pfnWrite */
2644	qedWrite,
2645	/* pfnFlush */
2646	qedFlush,
2647	/* pfnDiscard */
2648	NULL,
2649	/* pfnGetVersion */
2650	qedGetVersion,
2651	/* pfnGetSectorSize */
2652	qedGetSectorSize,
2653	/* pfnGetSize */
2654	qedGetSize,
2655	/* pfnGetFileSize */
2656	qedGetFileSize,
2657	/* pfnGetPCHSGeometry */
2658	qedGetPCHSGeometry,
2659	/* pfnSetPCHSGeometry */
2660	qedSetPCHSGeometry,
2661	/* pfnGetLCHSGeometry */
2662	qedGetLCHSGeometry,
2663	/* pfnSetLCHSGeometry */
2664	qedSetLCHSGeometry,
2665	/* pfnGetImageFlags */
2666	qedGetImageFlags,
2667	/* pfnGetOpenFlags */
2668	qedGetOpenFlags,
2669	/* pfnSetOpenFlags */
2670	qedSetOpenFlags,
2671	/* pfnGetComment */
2672	qedGetComment,
2673	/* pfnSetComment */
2674	qedSetComment,
2675	/* pfnGetUuid */
2676	qedGetUuid,
2677	/* pfnSetUuid */
2678	qedSetUuid,
2679	/* pfnGetModificationUuid */
2680	qedGetModificationUuid,
2681	/* pfnSetModificationUuid */
2682	qedSetModificationUuid,
2683	/* pfnGetParentUuid */
2684	qedGetParentUuid,
2685	/* pfnSetParentUuid */
2686	qedSetParentUuid,
2687	/* pfnGetParentModificationUuid */
2688	qedGetParentModificationUuid,
2689	/* pfnSetParentModificationUuid */
2690	qedSetParentModificationUuid,
2691	/* pfnDump */
2692	qedDump,
2693	/* pfnGetTimeStamp */
2694	NULL,
2695	/* pfnGetParentTimeStamp */
2696	NULL,
2697	/* pfnSetParentTimeStamp */
2698	NULL,
2699	/* pfnGetParentFilename */
2700	qedGetParentFilename,
2701	/* pfnSetParentFilename */
2702	qedSetParentFilename,
2703	/* pfnComposeLocation */
2704	genericFileComposeLocation,
2705	/* pfnComposeName */
2706	genericFileComposeName,
2707	/* pfnCompact */
2708	NULL,
2709	/* pfnResize */
2710	qedResize,
2711	/* pfnRepair */
2712	NULL,
2713	/* pfnTraverseMetadata */
2714	NULL
2715	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 56993

Download in other formats: