QED.cpp@ 62746

Last change on this file since 62746 was 62746, checked in by vboxsync, 8 years ago
Storage: warnings.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 89.9 KB

Line
1	/* $Id: QED.cpp 62746 2016-07-30 15:59:55Z vboxsync $ */
2	/** @file
3	* QED - QED Disk image.
4	*/
5
6	/*
7	* Copyright (C) 2011-2016 Oracle Corporation
8	*
9	* This file is part of VirtualBox Open Source Edition (OSE), as
10	* available from http://www.virtualbox.org. This file is free software;
11	* you can redistribute it and/or modify it under the terms of the GNU
12	* General Public License (GPL) as published by the Free Software
13	* Foundation, in version 2 as it comes in the "COPYING" file of the
14	* VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15	* hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16	*/
17
18
19	/*********************************************************************************************************************************
20	* Header Files *
21	*********************************************************************************************************************************/
22	#define LOG_GROUP LOG_GROUP_VD_QED
23	#include <VBox/vd-plugin.h>
24	#include <VBox/err.h>
25
26	#include <VBox/log.h>
27	#include <iprt/asm.h>
28	#include <iprt/assert.h>
29	#include <iprt/string.h>
30	#include <iprt/alloc.h>
31	#include <iprt/path.h>
32	#include <iprt/list.h>
33
34	#include "VDBackends.h"
35
36	/**
37	* The QED backend implements support for the qemu enhanced disk format (short QED)
38	* The specification for the format is available under http://wiki.qemu.org/Features/QED/Specification
39	*
40	* Missing things to implement:
41	* - compaction
42	* - resizing which requires block relocation (very rare case)
43	*/
44
45
46	/*********************************************************************************************************************************
47	* Structures in a QED image, little endian *
48	*********************************************************************************************************************************/
49
50	#pragma pack(1)
51	typedef struct QedHeader
52	{
53	/** Magic value. */
54	uint32_t u32Magic;
55	/** Cluster size in bytes. */
56	uint32_t u32ClusterSize;
57	/** Size of L1 and L2 tables in clusters. */
58	uint32_t u32TableSize;
59	/** size of this header structure in clusters. */
60	uint32_t u32HeaderSize;
61	/** Features used for the image. */
62	uint64_t u64FeatureFlags;
63	/** Compatibility features used for the image. */
64	uint64_t u64CompatFeatureFlags;
65	/** Self resetting feature bits. */
66	uint64_t u64AutoresetFeatureFlags;
67	/** Offset of the L1 table in bytes. */
68	uint64_t u64OffL1Table;
69	/** Logical image size as seen by the guest. */
70	uint64_t u64Size;
71	/** Offset of the backing filename in bytes. */
72	uint32_t u32OffBackingFilename;
73	/** Size of the backing filename. */
74	uint32_t u32BackingFilenameSize;
75	} QedHeader;
76	#pragma pack()
77	/** Pointer to a on disk QED header. */
78	typedef QedHeader *PQedHeader;
79
80	/** QED magic value. */
81	#define QED_MAGIC UINT32_C(0x00444551) /* QED\0 */
82	/** Cluster size minimum. */
83	#define QED_CLUSTER_SIZE_MIN RT_BIT(12)
84	/** Cluster size maximum. */
85	#define QED_CLUSTER_SIZE_MAX RT_BIT(26)
86	/** L1 and L2 Table size minimum. */
87	#define QED_TABLE_SIZE_MIN 1
88	/** L1 and L2 Table size maximum. */
89	#define QED_TABLE_SIZE_MAX 16
90
91	/** QED default cluster size when creating an image. */
92	#define QED_CLUSTER_SIZE_DEFAULT (64 * _1K)
93	/** The default table size in clusters. */
94	#define QED_TABLE_SIZE_DEFAULT 4
95
96	/** Feature flags.
97	* @{
98	*/
99	/** Image uses a backing file to provide data for unallocated clusters. */
100	#define QED_FEATURE_BACKING_FILE RT_BIT_64(0)
101	/** Image needs checking before use. */
102	#define QED_FEATURE_NEED_CHECK RT_BIT_64(1)
103	/** Don't probe for format of the backing file, treat as raw image. */
104	#define QED_FEATURE_BACKING_FILE_NO_PROBE RT_BIT_64(2)
105	/** Mask of valid features. */
106	#define QED_FEATURE_MASK (QED_FEATURE_BACKING_FILE \| QED_FEATURE_NEED_CHECK \| QED_FEATURE_BACKING_FILE_NO_PROBE)
107	/** @} */
108
109	/** Compatibility feature flags.
110	* @{
111	*/
112	/** Mask of valid compatibility features. */
113	#define QED_COMPAT_FEATURE_MASK (0)
114	/** @} */
115
116	/** Autoreset feature flags.
117	* @{
118	*/
119	/** Mask of valid autoreset features. */
120	#define QED_AUTORESET_FEATURE_MASK (0)
121	/** @} */
122
123
124	/*********************************************************************************************************************************
125	* Constants And Macros, Structures and Typedefs *
126	*********************************************************************************************************************************/
127
128	/**
129	* QED L2 cache entry.
130	*/
131	typedef struct QEDL2CACHEENTRY
132	{
133	/** List node for the search list. */
134	RTLISTNODE NodeSearch;
135	/** List node for the LRU list. */
136	RTLISTNODE NodeLru;
137	/** Reference counter. */
138	uint32_t cRefs;
139	/** The offset of the L2 table, used as search key. */
140	uint64_t offL2Tbl;
141	/** Pointer to the cached L2 table. */
142	uint64_t *paL2Tbl;
143	} QEDL2CACHEENTRY, *PQEDL2CACHEENTRY;
144
145	/** Maximum amount of memory the cache is allowed to use. */
146	#define QED_L2_CACHE_MEMORY_MAX (2*_1M)
147
148	/**
149	* QED image data structure.
150	*/
151	typedef struct QEDIMAGE
152	{
153	/** Image name. */
154	const char *pszFilename;
155	/** Storage handle. */
156	PVDIOSTORAGE pStorage;
157
158	/** Pointer to the per-disk VD interface list. */
159	PVDINTERFACE pVDIfsDisk;
160	/** Pointer to the per-image VD interface list. */
161	PVDINTERFACE pVDIfsImage;
162	/** Error interface. */
163	PVDINTERFACEERROR pIfError;
164	/** I/O interface. */
165	PVDINTERFACEIOINT pIfIo;
166
167	/** Open flags passed by VBoxHD layer. */
168	unsigned uOpenFlags;
169	/** Image flags defined during creation or determined during open. */
170	unsigned uImageFlags;
171	/** Total size of the image. */
172	uint64_t cbSize;
173	/** Physical geometry of this image. */
174	VDGEOMETRY PCHSGeometry;
175	/** Logical geometry of this image. */
176	VDGEOMETRY LCHSGeometry;
177
178	/** Filename of the backing file if any. */
179	char *pszBackingFilename;
180	/** Offset of the filename in the image. */
181	uint32_t offBackingFilename;
182	/** Size of the backing filename excluding \0. */
183	uint32_t cbBackingFilename;
184
185	/** Size of the image, multiple of clusters. */
186	uint64_t cbImage;
187	/** Cluster size in bytes. */
188	uint32_t cbCluster;
189	/** Number of entries in the L1 and L2 table. */
190	uint32_t cTableEntries;
191	/** Size of an L1 or L2 table rounded to the next cluster size. */
192	uint32_t cbTable;
193	/** Pointer to the L1 table. */
194	uint64_t *paL1Table;
195	/** Offset of the L1 table. */
196	uint64_t offL1Table;
197
198	/** Offset mask for a cluster. */
199	uint64_t fOffsetMask;
200	/** L1 table mask to get the L1 index. */
201	uint64_t fL1Mask;
202	/** Number of bits to shift to get the L1 index. */
203	uint32_t cL1Shift;
204	/** L2 table mask to get the L2 index. */
205	uint64_t fL2Mask;
206	/** Number of bits to shift to get the L2 index. */
207	uint32_t cL2Shift;
208
209	/** Memory occupied by the L2 table cache. */
210	size_t cbL2Cache;
211	/** The sorted L2 entry list used for searching. */
212	RTLISTNODE ListSearch;
213	/** The LRU L2 entry list used for eviction. */
214	RTLISTNODE ListLru;
215
216	} QEDIMAGE, *PQEDIMAGE;
217
218	/**
219	* State of the async cluster allocation.
220	*/
221	typedef enum QEDCLUSTERASYNCALLOCSTATE
222	{
223	/** Invalid. */
224	QEDCLUSTERASYNCALLOCSTATE_INVALID = 0,
225	/** L2 table allocation. */
226	QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC,
227	/** Link L2 table into L1. */
228	QEDCLUSTERASYNCALLOCSTATE_L2_LINK,
229	/** Allocate user data cluster. */
230	QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC,
231	/** Link user data cluster. */
232	QEDCLUSTERASYNCALLOCSTATE_USER_LINK,
233	/** 32bit blowup. */
234	QEDCLUSTERASYNCALLOCSTATE_32BIT_HACK = 0x7fffffff
235	} QEDCLUSTERASYNCALLOCSTATE, *PQEDCLUSTERASYNCALLOCSTATE;
236
237	/**
238	* Data needed to track async cluster allocation.
239	*/
240	typedef struct QEDCLUSTERASYNCALLOC
241	{
242	/** The state of the cluster allocation. */
243	QEDCLUSTERASYNCALLOCSTATE enmAllocState;
244	/** Old image size to rollback in case of an error. */
245	uint64_t cbImageOld;
246	/** L1 index to link if any. */
247	uint32_t idxL1;
248	/** L2 index to link, required in any case. */
249	uint32_t idxL2;
250	/** Start offset of the allocated cluster. */
251	uint64_t offClusterNew;
252	/** L2 cache entry if a L2 table is allocated. */
253	PQEDL2CACHEENTRY pL2Entry;
254	/** Number of bytes to write. */
255	size_t cbToWrite;
256	} QEDCLUSTERASYNCALLOC, *PQEDCLUSTERASYNCALLOC;
257
258
259	/*********************************************************************************************************************************
260	* Static Variables *
261	*********************************************************************************************************************************/
262
263	/** NULL-terminated array of supported file extensions. */
264	static const VDFILEEXTENSION s_aQedFileExtensions[] =
265	{
266	{"qed", VDTYPE_HDD},
267	{NULL, VDTYPE_INVALID}
268	};
269
270
271	/*********************************************************************************************************************************
272	* Internal Functions *
273	*********************************************************************************************************************************/
274
275	/**
276	* Converts the image header to the host endianess and performs basic checks.
277	*
278	* @returns Whether the given header is valid or not.
279	* @param pHeader Pointer to the header to convert.
280	*/
281	static bool qedHdrConvertToHostEndianess(PQedHeader pHeader)
282	{
283	pHeader->u32Magic = RT_LE2H_U32(pHeader->u32Magic);
284	pHeader->u32ClusterSize = RT_LE2H_U32(pHeader->u32ClusterSize);
285	pHeader->u32TableSize = RT_LE2H_U32(pHeader->u32TableSize);
286	pHeader->u32HeaderSize = RT_LE2H_U32(pHeader->u32HeaderSize);
287	pHeader->u64FeatureFlags = RT_LE2H_U64(pHeader->u64FeatureFlags);
288	pHeader->u64CompatFeatureFlags = RT_LE2H_U64(pHeader->u64CompatFeatureFlags);
289	pHeader->u64AutoresetFeatureFlags = RT_LE2H_U64(pHeader->u64AutoresetFeatureFlags);
290	pHeader->u64OffL1Table = RT_LE2H_U64(pHeader->u64OffL1Table);
291	pHeader->u64Size = RT_LE2H_U64(pHeader->u64Size);
292	pHeader->u32OffBackingFilename = RT_LE2H_U32(pHeader->u32OffBackingFilename);
293	pHeader->u32BackingFilenameSize = RT_LE2H_U32(pHeader->u32BackingFilenameSize);
294
295	if (RT_UNLIKELY(pHeader->u32Magic != QED_MAGIC))
296	return false;
297	if (RT_UNLIKELY( pHeader->u32ClusterSize < QED_CLUSTER_SIZE_MIN
298	\|\| pHeader->u32ClusterSize > QED_CLUSTER_SIZE_MAX))
299	return false;
300	if (RT_UNLIKELY( pHeader->u32TableSize < QED_TABLE_SIZE_MIN
301	\|\| pHeader->u32TableSize > QED_TABLE_SIZE_MAX))
302	return false;
303	if (RT_UNLIKELY(pHeader->u64Size % 512 != 0))
304	return false;
305
306	return true;
307	}
308
309	/**
310	* Creates a QED header from the given image state.
311	*
312	* @returns nothing.
313	* @param pImage Image instance data.
314	* @param pHeader Pointer to the header to convert.
315	*/
316	static void qedHdrConvertFromHostEndianess(PQEDIMAGE pImage, PQedHeader pHeader)
317	{
318	pHeader->u32Magic = RT_H2LE_U32(QED_MAGIC);
319	pHeader->u32ClusterSize = RT_H2LE_U32(pImage->cbCluster);
320	pHeader->u32TableSize = RT_H2LE_U32(pImage->cbTable / pImage->cbCluster);
321	pHeader->u32HeaderSize = RT_H2LE_U32(1);
322	pHeader->u64FeatureFlags = RT_H2LE_U64(pImage->pszBackingFilename ? QED_FEATURE_BACKING_FILE : UINT64_C(0));
323	pHeader->u64CompatFeatureFlags = RT_H2LE_U64(UINT64_C(0));
324	pHeader->u64AutoresetFeatureFlags = RT_H2LE_U64(UINT64_C(0));
325	pHeader->u64OffL1Table = RT_H2LE_U64(pImage->offL1Table);
326	pHeader->u64Size = RT_H2LE_U64(pImage->cbSize);
327	pHeader->u32OffBackingFilename = RT_H2LE_U32(pImage->offBackingFilename);
328	pHeader->u32BackingFilenameSize = RT_H2LE_U32(pImage->cbBackingFilename);
329	}
330
331	/**
332	* Convert table entries from little endian to host endianess.
333	*
334	* @returns nothing.
335	* @param paTbl Pointer to the table.
336	* @param cEntries Number of entries in the table.
337	*/
338	static void qedTableConvertToHostEndianess(uint64_t *paTbl, uint32_t cEntries)
339	{
340	while(cEntries-- > 0)
341	{
342	paTbl = RT_LE2H_U64(paTbl);
343	paTbl++;
344	}
345	}
346
347	/**
348	* Convert table entries from host to little endian format.
349	*
350	* @returns nothing.
351	* @param paTblImg Pointer to the table which will store the little endian table.
352	* @param paTbl The source table to convert.
353	* @param cEntries Number of entries in the table.
354	*/
355	static void qedTableConvertFromHostEndianess(uint64_t paTblImg, uint64_t paTbl,
356	uint32_t cEntries)
357	{
358	while(cEntries-- > 0)
359	{
360	paTblImg = RT_H2LE_U64(paTbl);
361	paTbl++;
362	paTblImg++;
363	}
364	}
365
366	/**
367	* Creates the L2 table cache.
368	*
369	* @returns VBox status code.
370	* @param pImage The image instance data.
371	*/
372	static int qedL2TblCacheCreate(PQEDIMAGE pImage)
373	{
374	pImage->cbL2Cache = 0;
375	RTListInit(&pImage->ListSearch);
376	RTListInit(&pImage->ListLru);
377
378	return VINF_SUCCESS;
379	}
380
381	/**
382	* Destroys the L2 table cache.
383	*
384	* @returns nothing.
385	* @param pImage The image instance data.
386	*/
387	static void qedL2TblCacheDestroy(PQEDIMAGE pImage)
388	{
389	PQEDL2CACHEENTRY pL2Entry = NULL;
390	PQEDL2CACHEENTRY pL2Next = NULL;
391
392	RTListForEachSafe(&pImage->ListSearch, pL2Entry, pL2Next, QEDL2CACHEENTRY, NodeSearch)
393	{
394	Assert(!pL2Entry->cRefs);
395
396	RTListNodeRemove(&pL2Entry->NodeSearch);
397	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
398	RTMemFree(pL2Entry);
399	}
400
401	pImage->cbL2Cache = 0;
402	RTListInit(&pImage->ListSearch);
403	RTListInit(&pImage->ListLru);
404	}
405
406	/**
407	* Returns the L2 table matching the given offset or NULL if none could be found.
408	*
409	* @returns Pointer to the L2 table cache entry or NULL.
410	* @param pImage The image instance data.
411	* @param offL2Tbl Offset of the L2 table to search for.
412	*/
413	static PQEDL2CACHEENTRY qedL2TblCacheRetain(PQEDIMAGE pImage, uint64_t offL2Tbl)
414	{
415	PQEDL2CACHEENTRY pL2Entry = NULL;
416
417	RTListForEach(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch)
418	{
419	if (pL2Entry->offL2Tbl == offL2Tbl)
420	break;
421	}
422
423	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
424	{
425	/* Update LRU list. */
426	RTListNodeRemove(&pL2Entry->NodeLru);
427	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
428	pL2Entry->cRefs++;
429	return pL2Entry;
430	}
431	else
432	return NULL;
433	}
434
435	/**
436	* Releases a L2 table cache entry.
437	*
438	* @returns nothing.
439	* @param pL2Entry The L2 cache entry.
440	*/
441	static void qedL2TblCacheEntryRelease(PQEDL2CACHEENTRY pL2Entry)
442	{
443	Assert(pL2Entry->cRefs > 0);
444	pL2Entry->cRefs--;
445	}
446
447	/**
448	* Allocates a new L2 table from the cache evicting old entries if required.
449	*
450	* @returns Pointer to the L2 cache entry or NULL.
451	* @param pImage The image instance data.
452	*/
453	static PQEDL2CACHEENTRY qedL2TblCacheEntryAlloc(PQEDIMAGE pImage)
454	{
455	PQEDL2CACHEENTRY pL2Entry = NULL;
456
457	if (pImage->cbL2Cache + pImage->cbTable <= QED_L2_CACHE_MEMORY_MAX)
458	{
459	/* Add a new entry. */
460	pL2Entry = (PQEDL2CACHEENTRY)RTMemAllocZ(sizeof(QEDL2CACHEENTRY));
461	if (pL2Entry)
462	{
463	pL2Entry->paL2Tbl = (uint64_t *)RTMemPageAllocZ(pImage->cbTable);
464	if (RT_UNLIKELY(!pL2Entry->paL2Tbl))
465	{
466	RTMemFree(pL2Entry);
467	pL2Entry = NULL;
468	}
469	else
470	{
471	pL2Entry->cRefs = 1;
472	pImage->cbL2Cache += pImage->cbTable;
473	}
474	}
475	}
476	else
477	{
478	/* Evict the last not in use entry and use it */
479	Assert(!RTListIsEmpty(&pImage->ListLru));
480
481	RTListForEachReverse(&pImage->ListLru, pL2Entry, QEDL2CACHEENTRY, NodeLru)
482	{
483	if (!pL2Entry->cRefs)
484	break;
485	}
486
487	if (!RTListNodeIsDummy(&pImage->ListSearch, pL2Entry, QEDL2CACHEENTRY, NodeSearch))
488	{
489	RTListNodeRemove(&pL2Entry->NodeSearch);
490	RTListNodeRemove(&pL2Entry->NodeLru);
491	pL2Entry->offL2Tbl = 0;
492	pL2Entry->cRefs = 1;
493	}
494	else
495	pL2Entry = NULL;
496	}
497
498	return pL2Entry;
499	}
500
501	/**
502	* Frees a L2 table cache entry.
503	*
504	* @returns nothing.
505	* @param pImage The image instance data.
506	* @param pL2Entry The L2 cache entry to free.
507	*/
508	static void qedL2TblCacheEntryFree(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
509	{
510	Assert(!pL2Entry->cRefs);
511	RTMemPageFree(pL2Entry->paL2Tbl, pImage->cbTable);
512	RTMemFree(pL2Entry);
513
514	pImage->cbL2Cache -= pImage->cbTable;
515	}
516
517	/**
518	* Inserts an entry in the L2 table cache.
519	*
520	* @returns nothing.
521	* @param pImage The image instance data.
522	* @param pL2Entry The L2 cache entry to insert.
523	*/
524	static void qedL2TblCacheEntryInsert(PQEDIMAGE pImage, PQEDL2CACHEENTRY pL2Entry)
525	{
526	PQEDL2CACHEENTRY pIt = NULL;
527
528	Assert(pL2Entry->offL2Tbl > 0);
529
530	/* Insert at the top of the LRU list. */
531	RTListPrepend(&pImage->ListLru, &pL2Entry->NodeLru);
532
533	if (RTListIsEmpty(&pImage->ListSearch))
534	{
535	RTListAppend(&pImage->ListSearch, &pL2Entry->NodeSearch);
536	}
537	else
538	{
539	/* Insert into search list. */
540	pIt = RTListGetFirst(&pImage->ListSearch, QEDL2CACHEENTRY, NodeSearch);
541	if (pIt->offL2Tbl > pL2Entry->offL2Tbl)
542	RTListPrepend(&pImage->ListSearch, &pL2Entry->NodeSearch);
543	else
544	{
545	bool fInserted = false;
546
547	RTListForEach(&pImage->ListSearch, pIt, QEDL2CACHEENTRY, NodeSearch)
548	{
549	Assert(pIt->offL2Tbl != pL2Entry->offL2Tbl);
550	if (pIt->offL2Tbl < pL2Entry->offL2Tbl)
551	{
552	RTListNodeInsertAfter(&pIt->NodeSearch, &pL2Entry->NodeSearch);
553	fInserted = true;
554	break;
555	}
556	}
557	Assert(fInserted);
558	}
559	}
560	}
561
562	/**
563	* Fetches the L2 from the given offset trying the LRU cache first and
564	* reading it from the image after a cache miss.
565	*
566	* @returns VBox status code.
567	* @param pImage Image instance data.
568	* @param offL2Tbl The offset of the L2 table in the image.
569	* @param ppL2Entry Where to store the L2 table on success.
570	*/
571	static int qedL2TblCacheFetch(PQEDIMAGE pImage, uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
572	{
573	int rc = VINF_SUCCESS;
574
575	LogFlowFunc(("pImage=%#p offL2Tbl=%llu ppL2Entry=%#p\n", pImage, offL2Tbl, ppL2Entry));
576
577	/* Try to fetch the L2 table from the cache first. */
578	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
579	if (!pL2Entry)
580	{
581	LogFlowFunc(("Reading L2 table from image\n"));
582	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
583
584	if (pL2Entry)
585	{
586	/* Read from the image. */
587	pL2Entry->offL2Tbl = offL2Tbl;
588	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, offL2Tbl,
589	pL2Entry->paL2Tbl, pImage->cbTable);
590	if (RT_SUCCESS(rc))
591	{
592	#if defined(RT_BIG_ENDIAN)
593	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
594	#endif
595	qedL2TblCacheEntryInsert(pImage, pL2Entry);
596	}
597	else
598	{
599	qedL2TblCacheEntryRelease(pL2Entry);
600	qedL2TblCacheEntryFree(pImage, pL2Entry);
601	}
602	}
603	else
604	rc = VERR_NO_MEMORY;
605	}
606
607	if (RT_SUCCESS(rc))
608	*ppL2Entry = pL2Entry;
609
610	LogFlowFunc(("returns rc=%Rrc\n", rc));
611	return rc;
612	}
613
614	/**
615	* Fetches the L2 from the given offset trying the LRU cache first and
616	* reading it from the image after a cache miss - version for async I/O.
617	*
618	* @returns VBox status code.
619	* @param pImage Image instance data.
620	* @param pIoCtx The I/O context.
621	* @param offL2Tbl The offset of the L2 table in the image.
622	* @param ppL2Entry Where to store the L2 table on success.
623	*/
624	static int qedL2TblCacheFetchAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
625	uint64_t offL2Tbl, PQEDL2CACHEENTRY *ppL2Entry)
626	{
627	int rc = VINF_SUCCESS;
628
629	/* Try to fetch the L2 table from the cache first. */
630	PQEDL2CACHEENTRY pL2Entry = qedL2TblCacheRetain(pImage, offL2Tbl);
631	if (!pL2Entry)
632	{
633	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
634
635	if (pL2Entry)
636	{
637	/* Read from the image. */
638	PVDMETAXFER pMetaXfer;
639
640	pL2Entry->offL2Tbl = offL2Tbl;
641	rc = vdIfIoIntFileReadMeta(pImage->pIfIo, pImage->pStorage,
642	offL2Tbl, pL2Entry->paL2Tbl,
643	pImage->cbTable, pIoCtx,
644	&pMetaXfer, NULL, NULL);
645	if (RT_SUCCESS(rc))
646	{
647	vdIfIoIntMetaXferRelease(pImage->pIfIo, pMetaXfer);
648	#if defined(RT_BIG_ENDIAN)
649	qedTableConvertToHostEndianess(pL2Entry->paL2Tbl, pImage->cTableEntries);
650	#endif
651	qedL2TblCacheEntryInsert(pImage, pL2Entry);
652	}
653	else
654	{
655	qedL2TblCacheEntryRelease(pL2Entry);
656	qedL2TblCacheEntryFree(pImage, pL2Entry);
657	}
658	}
659	else
660	rc = VERR_NO_MEMORY;
661	}
662
663	if (RT_SUCCESS(rc))
664	*ppL2Entry = pL2Entry;
665
666	return rc;
667	}
668
669	/**
670	* Return power of 2 or 0 if num error.
671	*
672	* @returns The power of 2 or 0 if the given number is not a power of 2.
673	* @param u32 The number.
674	*/
675	static uint32_t qedGetPowerOfTwo(uint32_t u32)
676	{
677	if (u32 == 0)
678	return 0;
679	uint32_t uPower2 = 0;
680	while ((u32 & 1) == 0)
681	{
682	u32 >>= 1;
683	uPower2++;
684	}
685	return u32 == 1 ? uPower2 : 0;
686	}
687
688	/**
689	* Sets the L1, L2 and offset bitmasks and L1 and L2 bit shift members.
690	*
691	* @returns nothing.
692	* @param pImage The image instance data.
693	*/
694	static void qedTableMasksInit(PQEDIMAGE pImage)
695	{
696	uint32_t cClusterBits, cTableBits;
697
698	cClusterBits = qedGetPowerOfTwo(pImage->cbCluster);
699	cTableBits = qedGetPowerOfTwo(pImage->cTableEntries);
700
701	Assert(cClusterBits + 2 * cTableBits <= 64);
702
703	pImage->fOffsetMask = ((uint64_t)pImage->cbCluster - 1);
704	pImage->fL2Mask = ((uint64_t)pImage->cTableEntries - 1) << cClusterBits;
705	pImage->cL2Shift = cClusterBits;
706	pImage->fL1Mask = ((uint64_t)pImage->cTableEntries - 1) << (cClusterBits + cTableBits);
707	pImage->cL1Shift = cClusterBits + cTableBits;
708	}
709
710	/**
711	* Converts a given logical offset into the
712	*
713	* @returns nothing.
714	* @param pImage The image instance data.
715	* @param off The logical offset to convert.
716	* @param pidxL1 Where to store the index in the L1 table on success.
717	* @param pidxL2 Where to store the index in the L2 table on success.
718	* @param poffCluster Where to store the offset in the cluster on success.
719	*/
720	DECLINLINE(void) qedConvertLogicalOffset(PQEDIMAGE pImage, uint64_t off, uint32_t *pidxL1,
721	uint32_t pidxL2, uint32_t poffCluster)
722	{
723	AssertPtr(pidxL1);
724	AssertPtr(pidxL2);
725	AssertPtr(poffCluster);
726
727	*poffCluster = off & pImage->fOffsetMask;
728	*pidxL1 = (off & pImage->fL1Mask) >> pImage->cL1Shift;
729	*pidxL2 = (off & pImage->fL2Mask) >> pImage->cL2Shift;
730	}
731
732	/**
733	* Converts Cluster size to a byte size.
734	*
735	* @returns Number of bytes derived from the given number of clusters.
736	* @param pImage The image instance data.
737	* @param cClusters The clusters to convert.
738	*/
739	DECLINLINE(uint64_t) qedCluster2Byte(PQEDIMAGE pImage, uint64_t cClusters)
740	{
741	return cClusters * pImage->cbCluster;
742	}
743
744	/**
745	* Converts number of bytes to cluster size rounding to the next cluster.
746	*
747	* @returns Number of bytes derived from the given number of clusters.
748	* @param pImage The image instance data.
749	* @param cb Number of bytes to convert.
750	*/
751	DECLINLINE(uint64_t) qedByte2Cluster(PQEDIMAGE pImage, uint64_t cb)
752	{
753	return cb / pImage->cbCluster + (cb % pImage->cbCluster ? 1 : 0);
754	}
755
756	/**
757	* Allocates a new cluster in the image.
758	*
759	* @returns The start offset of the new cluster in the image.
760	* @param pImage The image instance data.
761	* @param cCLusters Number of clusters to allocate.
762	*/
763	DECLINLINE(uint64_t) qedClusterAllocate(PQEDIMAGE pImage, uint32_t cClusters)
764	{
765	uint64_t offCluster;
766
767	offCluster = pImage->cbImage;
768	pImage->cbImage += cClusters*pImage->cbCluster;
769
770	return offCluster;
771	}
772
773	/**
774	* Returns the real image offset for a given cluster or an error if the cluster is not
775	* yet allocated.
776	*
777	* @returns VBox status code.
778	* VERR_VD_BLOCK_FREE if the cluster is not yet allocated.
779	* @param pImage The image instance data.
780	* @param pIoCtx The I/O context.
781	* @param idxL1 The L1 index.
782	* @param idxL2 The L2 index.
783	* @param offCluster Offset inside the cluster.
784	* @param poffImage Where to store the image offset on success;
785	*/
786	static int qedConvertToImageOffset(PQEDIMAGE pImage, PVDIOCTX pIoCtx,
787	uint32_t idxL1, uint32_t idxL2,
788	uint32_t offCluster, uint64_t *poffImage)
789	{
790	int rc = VERR_VD_BLOCK_FREE;
791
792	AssertReturn(idxL1 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
793	AssertReturn(idxL2 < pImage->cTableEntries, VERR_INVALID_PARAMETER);
794
795	if (pImage->paL1Table[idxL1])
796	{
797	PQEDL2CACHEENTRY pL2Entry;
798
799	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
800	&pL2Entry);
801	if (RT_SUCCESS(rc))
802	{
803	/* Get real file offset. */
804	if (pL2Entry->paL2Tbl[idxL2])
805	*poffImage = pL2Entry->paL2Tbl[idxL2] + offCluster;
806	else
807	rc = VERR_VD_BLOCK_FREE;
808
809	qedL2TblCacheEntryRelease(pL2Entry);
810	}
811	}
812
813	return rc;
814	}
815
816
817	/**
818	* Internal. Flush image data to disk.
819	*/
820	static int qedFlushImage(PQEDIMAGE pImage)
821	{
822	int rc = VINF_SUCCESS;
823
824	if ( pImage->pStorage
825	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
826	{
827	QedHeader Header;
828
829	Assert(!(pImage->cbTable % pImage->cbCluster));
830	#if defined(RT_BIG_ENDIAN)
831	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
832	if (paL1TblImg)
833	{
834	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
835	pImage->cTableEntries);
836	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
837	pImage->offL1Table, paL1TblImg,
838	pImage->cbTable);
839	RTMemFree(paL1TblImg);
840	}
841	else
842	rc = VERR_NO_MEMORY;
843	#else
844	/* Write L1 table directly. */
845	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, pImage->offL1Table,
846	pImage->paL1Table, pImage->cbTable);
847	#endif
848	if (RT_SUCCESS(rc))
849	{
850	/* Write header. */
851	qedHdrConvertFromHostEndianess(pImage, &Header);
852	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage, 0, &Header,
853	sizeof(Header));
854	if (RT_SUCCESS(rc))
855	rc = vdIfIoIntFileFlushSync(pImage->pIfIo, pImage->pStorage);
856	}
857	}
858
859	return rc;
860	}
861
862	/**
863	* Flush image data to disk - version for async I/O.
864	*
865	* @returns VBox status code.
866	* @param pImage The image instance data.
867	* @param pIoCtx The I/o context
868	*/
869	static int qedFlushImageAsync(PQEDIMAGE pImage, PVDIOCTX pIoCtx)
870	{
871	int rc = VINF_SUCCESS;
872
873	if ( pImage->pStorage
874	&& !(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
875	{
876	QedHeader Header;
877
878	Assert(!(pImage->cbTable % pImage->cbCluster));
879	#if defined(RT_BIG_ENDIAN)
880	uint64_t paL1TblImg = (uint64_t )RTMemAllocZ(pImage->cbTable);
881	if (paL1TblImg)
882	{
883	qedTableConvertFromHostEndianess(paL1TblImg, pImage->paL1Table,
884	pImage->cTableEntries);
885	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
886	pImage->offL1Table, paL1TblImg,
887	pImage->cbTable, pIoCtx, NULL, NULL);
888	RTMemFree(paL1TblImg);
889	}
890	else
891	rc = VERR_NO_MEMORY;
892	#else
893	/* Write L1 table directly. */
894	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
895	pImage->offL1Table, pImage->paL1Table,
896	pImage->cbTable, pIoCtx, NULL, NULL);
897	#endif
898	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
899	{
900	/* Write header. */
901	qedHdrConvertFromHostEndianess(pImage, &Header);
902	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
903	0, &Header, sizeof(Header),
904	pIoCtx, NULL, NULL);
905	if (RT_SUCCESS(rc) \|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
906	rc = vdIfIoIntFileFlush(pImage->pIfIo, pImage->pStorage,
907	pIoCtx, NULL, NULL);
908	}
909	}
910
911	return rc;
912	}
913
914	/**
915	* Checks whether the given cluster offset is valid.
916	*
917	* @returns Whether the given cluster offset is valid.
918	* @param offCluster The table offset to check.
919	* @param cbFile The real file size of the image.
920	* @param cbCluster The cluster size in bytes.
921	*/
922	DECLINLINE(bool) qedIsClusterOffsetValid(uint64_t offCluster, uint64_t cbFile, size_t cbCluster)
923	{
924	return (offCluster <= cbFile - cbCluster)
925	&& !(offCluster & (cbCluster - 1));
926	}
927
928	/**
929	* Checks whether the given table offset is valid.
930	*
931	* @returns Whether the given table offset is valid.
932	* @param offTbl The table offset to check.
933	* @param cbFile The real file size of the image.
934	* @param cbTable The table size in bytes.
935	* @param cbCluster The cluster size in bytes.
936	*/
937	DECLINLINE(bool) qedIsTblOffsetValid(uint64_t offTbl, uint64_t cbFile, size_t cbTable, size_t cbCluster)
938	{
939	return (offTbl <= cbFile - cbTable)
940	&& !(offTbl & (cbCluster - 1));
941	}
942
943	/**
944	* Sets the specified range in the cluster bitmap checking whether any of the clusters is already
945	* used before.
946	*
947	* @returns Whether the range was clear and is set now.
948	* @param pvClusterBitmap The cluster bitmap to use.
949	* @param offClusterStart The first cluster to check and set.
950	* @param offClusterEnd The first cluster to not check and set anymore.
951	*/
952	static bool qedClusterBitmapCheckAndSet(void *pvClusterBitmap, uint32_t offClusterStart, uint32_t offClusterEnd)
953	{
954	for (uint32_t offCluster = offClusterStart; offCluster < offClusterEnd; offCluster++)
955	if (ASMBitTest(pvClusterBitmap, offCluster))
956	return false;
957
958	ASMBitSetRange(pvClusterBitmap, offClusterStart, offClusterEnd);
959	return true;
960	}
961
962	/**
963	* Checks the given image for consistency, usually called when the
964	* QED_FEATURE_NEED_CHECK bit is set.
965	*
966	* @returns VBox status code.
967	* @retval VINF_SUCCESS when the image can be accessed.
968	* @param pImage The image instance data.
969	* @param pHeader The header to use for checking.
970	*
971	* @note It is not required that the image state is fully initialized Only
972	* The I/O interface and storage handle need to be valid.
973	* @note The header must be converted to the host CPU endian format already
974	* and should be validated already.
975	*/
976	static int qedCheckImage(PQEDIMAGE pImage, PQedHeader pHeader)
977	{
978	uint64_t cbFile;
979	uint32_t cbTable;
980	uint32_t cTableEntries;
981	uint64_t *paL1Tbl = NULL;
982	uint64_t *paL2Tbl = NULL;
983	void *pvClusterBitmap = NULL;
984	uint32_t offClusterStart;
985	int rc = VINF_SUCCESS;
986
987	pImage->cbCluster = pHeader->u32ClusterSize;
988	cbTable = pHeader->u32TableSize * pHeader->u32ClusterSize;
989	cTableEntries = cbTable / sizeof(uint64_t);
990
991	do
992	{
993	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
994	if (RT_FAILURE(rc))
995	{
996	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
997	N_("Qed: Querying the file size of image '%s' failed"),
998	pImage->pszFilename);
999	break;
1000	}
1001
1002	/* Allocate L1 table. */
1003	paL1Tbl = (uint64_t *)RTMemAllocZ(cbTable);
1004	if (!paL1Tbl)
1005	{
1006	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1007	N_("Qed: Allocating memory for the L1 table for image '%s' failed"),
1008	pImage->pszFilename);
1009	break;
1010	}
1011
1012	paL2Tbl = (uint64_t *)RTMemAllocZ(cbTable);
1013	if (!paL2Tbl)
1014	{
1015	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1016	N_("Qed: Allocating memory for the L2 table for image '%s' failed"),
1017	pImage->pszFilename);
1018	break;
1019	}
1020
1021	pvClusterBitmap = RTMemAllocZ(cbFile / pHeader->u32ClusterSize / 8);
1022	if (!pvClusterBitmap)
1023	{
1024	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1025	N_("Qed: Allocating memory for the cluster bitmap for image '%s' failed"),
1026	pImage->pszFilename);
1027	break;
1028	}
1029
1030	/* Validate L1 table offset. */
1031	if (!qedIsTblOffsetValid(pHeader->u64OffL1Table, cbFile, cbTable, pHeader->u32ClusterSize))
1032	{
1033	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1034	N_("Qed: L1 table offset of image '%s' is corrupt (%llu)"),
1035	pImage->pszFilename, pHeader->u64OffL1Table);
1036	break;
1037	}
1038
1039	/* Read L1 table. */
1040	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1041	pHeader->u64OffL1Table, paL1Tbl, cbTable);
1042	if (RT_FAILURE(rc))
1043	{
1044	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1045	N_("Qed: Reading the L1 table from image '%s' failed"),
1046	pImage->pszFilename);
1047	break;
1048	}
1049
1050	/* Mark the L1 table in cluster bitmap. */
1051	ASMBitSet(pvClusterBitmap, 0); /* Header is always in cluster 0. */
1052	offClusterStart = qedByte2Cluster(pImage, pHeader->u64OffL1Table);
1053	bool fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1054	Assert(fSet);
1055
1056	/* Scan the L1 and L2 tables for invalid entries. */
1057	qedTableConvertToHostEndianess(paL1Tbl, cTableEntries);
1058
1059	for (unsigned iL1 = 0; iL1 < cTableEntries; iL1++)
1060	{
1061	if (!paL1Tbl[iL1])
1062	continue; /* Skip unallocated clusters. */
1063
1064	if (!qedIsTblOffsetValid(paL1Tbl[iL1], cbFile, cbTable, pHeader->u32ClusterSize))
1065	{
1066	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1067	N_("Qed: Entry %d of the L1 table from image '%s' is invalid (%llu)"),
1068	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1069	break;
1070	}
1071
1072	/* Now check that the clusters are not allocated already. */
1073	offClusterStart = qedByte2Cluster(pImage, paL1Tbl[iL1]);
1074	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + pHeader->u32TableSize);
1075	if (!fSet)
1076	{
1077	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1078	N_("Qed: Entry %d of the L1 table from image '%s' points to a already used cluster (%llu)"),
1079	iL1, pImage->pszFilename, paL1Tbl[iL1]);
1080	break;
1081	}
1082
1083	/* Read the linked L2 table and check it. */
1084	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1085	paL1Tbl[iL1], paL2Tbl, cbTable);
1086	if (RT_FAILURE(rc))
1087	{
1088	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1089	N_("Qed: Reading the L2 table from image '%s' failed"),
1090	pImage->pszFilename);
1091	break;
1092	}
1093
1094	/* Check all L2 entries. */
1095	for (unsigned iL2 = 0; iL2 < cTableEntries; iL2++)
1096	{
1097	if (paL2Tbl[iL2])
1098	continue; /* Skip unallocated clusters. */
1099
1100	if (!qedIsClusterOffsetValid(paL2Tbl[iL2], cbFile, pHeader->u32ClusterSize))
1101	{
1102	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1103	N_("Qed: Entry %d of the L2 table from image '%s' is invalid (%llu)"),
1104	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1105	break;
1106	}
1107
1108	/* Now check that the clusters are not allocated already. */
1109	offClusterStart = qedByte2Cluster(pImage, paL2Tbl[iL2]);
1110	fSet = qedClusterBitmapCheckAndSet(pvClusterBitmap, offClusterStart, offClusterStart + 1);
1111	if (!fSet)
1112	{
1113	rc = vdIfError(pImage->pIfError, VERR_VD_GEN_INVALID_HEADER, RT_SRC_POS,
1114	N_("Qed: Entry %d of the L2 table from image '%s' points to a already used cluster (%llu)"),
1115	iL2, pImage->pszFilename, paL2Tbl[iL2]);
1116	break;
1117	}
1118	}
1119	}
1120	} while(0);
1121
1122	if (paL1Tbl)
1123	RTMemFree(paL1Tbl);
1124	if (paL2Tbl)
1125	RTMemFree(paL2Tbl);
1126	if (pvClusterBitmap)
1127	RTMemFree(pvClusterBitmap);
1128
1129	return rc;
1130	}
1131
1132	/**
1133	* Internal. Free all allocated space for representing an image except pImage,
1134	* and optionally delete the image from disk.
1135	*/
1136	static int qedFreeImage(PQEDIMAGE pImage, bool fDelete)
1137	{
1138	int rc = VINF_SUCCESS;
1139
1140	/* Freeing a never allocated image (e.g. because the open failed) is
1141	* not signalled as an error. After all nothing bad happens. */
1142	if (pImage)
1143	{
1144	if (pImage->pStorage)
1145	{
1146	/* No point updating the file that is deleted anyway. */
1147	if (!fDelete)
1148	qedFlushImage(pImage);
1149
1150	rc = vdIfIoIntFileClose(pImage->pIfIo, pImage->pStorage);
1151	pImage->pStorage = NULL;
1152	}
1153
1154	if (pImage->paL1Table)
1155	RTMemFree(pImage->paL1Table);
1156
1157	if (pImage->pszBackingFilename)
1158	{
1159	RTMemFree(pImage->pszBackingFilename);
1160	pImage->pszBackingFilename = NULL;
1161	}
1162
1163	qedL2TblCacheDestroy(pImage);
1164
1165	if (fDelete && pImage->pszFilename)
1166	vdIfIoIntFileDelete(pImage->pIfIo, pImage->pszFilename);
1167	}
1168
1169	LogFlowFunc(("returns %Rrc\n", rc));
1170	return rc;
1171	}
1172
1173	/**
1174	* Internal: Open an image, constructing all necessary data structures.
1175	*/
1176	static int qedOpenImage(PQEDIMAGE pImage, unsigned uOpenFlags)
1177	{
1178	int rc;
1179
1180	pImage->uOpenFlags = uOpenFlags;
1181
1182	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1183	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1184	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1185
1186	/*
1187	* Create the L2 cache before opening the image so we can call qedFreeImage()
1188	* even if opening the image file fails.
1189	*/
1190	rc = qedL2TblCacheCreate(pImage);
1191	if (RT_FAILURE(rc))
1192	{
1193	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1194	N_("Qed: Creating the L2 table cache for image '%s' failed"),
1195	pImage->pszFilename);
1196
1197	goto out;
1198	}
1199
1200	/*
1201	* Open the image.
1202	*/
1203	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename,
1204	VDOpenFlagsToFileOpenFlags(uOpenFlags,
1205	false /* fCreate */),
1206	&pImage->pStorage);
1207	if (RT_FAILURE(rc))
1208	{
1209	/* Do NOT signal an appropriate error here, as the VD layer has the
1210	* choice of retrying the open if it failed. */
1211	goto out;
1212	}
1213
1214	uint64_t cbFile;
1215	QedHeader Header;
1216	rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
1217	if (RT_FAILURE(rc))
1218	goto out;
1219	if (cbFile > sizeof(Header))
1220	{
1221	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage, 0, &Header, sizeof(Header));
1222	if ( RT_SUCCESS(rc)
1223	&& qedHdrConvertToHostEndianess(&Header))
1224	{
1225	if ( !(Header.u64FeatureFlags & ~QED_FEATURE_MASK)
1226	&& !(Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE_NO_PROBE))
1227	{
1228	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1229	{
1230	/* Image needs checking. */
1231	if (!(uOpenFlags & VD_OPEN_FLAGS_READONLY))
1232	rc = qedCheckImage(pImage, &Header);
1233	else
1234	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1235	N_("Qed: Image '%s' needs checking but is opened readonly"),
1236	pImage->pszFilename);
1237	}
1238
1239	if ( RT_SUCCESS(rc)
1240	&& (Header.u64FeatureFlags & QED_FEATURE_BACKING_FILE))
1241	{
1242	/* Load backing filename from image. */
1243	pImage->pszBackingFilename = (char )RTMemAllocZ(Header.u32BackingFilenameSize + 1); / +1 for \0 terminator. */
1244	if (pImage->pszBackingFilename)
1245	{
1246	pImage->cbBackingFilename = Header.u32BackingFilenameSize;
1247	pImage->offBackingFilename = Header.u32OffBackingFilename;
1248	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1249	Header.u32OffBackingFilename, pImage->pszBackingFilename,
1250	Header.u32BackingFilenameSize);
1251	}
1252	else
1253	rc = VERR_NO_MEMORY;
1254	}
1255
1256	if (RT_SUCCESS(rc))
1257	{
1258	pImage->cbImage = cbFile;
1259	pImage->cbCluster = Header.u32ClusterSize;
1260	pImage->cbTable = Header.u32TableSize * pImage->cbCluster;
1261	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1262	pImage->offL1Table = Header.u64OffL1Table;
1263	pImage->cbSize = Header.u64Size;
1264	qedTableMasksInit(pImage);
1265
1266	/* Allocate L1 table. */
1267	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1268	if (pImage->paL1Table)
1269	{
1270	/* Read from the image. */
1271	rc = vdIfIoIntFileReadSync(pImage->pIfIo, pImage->pStorage,
1272	pImage->offL1Table, pImage->paL1Table,
1273	pImage->cbTable);
1274	if (RT_SUCCESS(rc))
1275	{
1276	qedTableConvertToHostEndianess(pImage->paL1Table, pImage->cTableEntries);
1277
1278	/* If the consistency check succeeded, clear the flag by flushing the image. */
1279	if (Header.u64FeatureFlags & QED_FEATURE_NEED_CHECK)
1280	rc = qedFlushImage(pImage);
1281	}
1282	else
1283	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS,
1284	N_("Qed: Reading the L1 table for image '%s' failed"),
1285	pImage->pszFilename);
1286	}
1287	else
1288	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS,
1289	N_("Qed: Out of memory allocating L1 table for image '%s'"),
1290	pImage->pszFilename);
1291	}
1292	}
1293	else
1294	rc = vdIfError(pImage->pIfError, VERR_NOT_SUPPORTED, RT_SRC_POS,
1295	N_("Qed: The image '%s' makes use of unsupported features"),
1296	pImage->pszFilename);
1297	}
1298	else if (RT_SUCCESS(rc))
1299	rc = VERR_VD_GEN_INVALID_HEADER;
1300	}
1301	else
1302	rc = VERR_VD_GEN_INVALID_HEADER;
1303
1304	out:
1305	if (RT_FAILURE(rc))
1306	qedFreeImage(pImage, false);
1307	return rc;
1308	}
1309
1310	/**
1311	* Internal: Create a qed image.
1312	*/
1313	static int qedCreateImage(PQEDIMAGE pImage, uint64_t cbSize,
1314	unsigned uImageFlags, const char *pszComment,
1315	PCVDGEOMETRY pPCHSGeometry,
1316	PCVDGEOMETRY pLCHSGeometry, unsigned uOpenFlags,
1317	PFNVDPROGRESS pfnProgress, void *pvUser,
1318	unsigned uPercentStart, unsigned uPercentSpan)
1319	{
1320	RT_NOREF1(pszComment);
1321	int rc;
1322	int32_t fOpen;
1323
1324	if (uImageFlags & VD_IMAGE_FLAGS_FIXED)
1325	{
1326	rc = vdIfError(pImage->pIfError, VERR_VD_INVALID_TYPE, RT_SRC_POS, N_("Qed: cannot create fixed image '%s'"), pImage->pszFilename);
1327	goto out;
1328	}
1329
1330	pImage->uOpenFlags = uOpenFlags & ~VD_OPEN_FLAGS_READONLY;
1331	pImage->uImageFlags = uImageFlags;
1332	pImage->PCHSGeometry = *pPCHSGeometry;
1333	pImage->LCHSGeometry = *pLCHSGeometry;
1334
1335	pImage->pIfError = VDIfErrorGet(pImage->pVDIfsDisk);
1336	pImage->pIfIo = VDIfIoIntGet(pImage->pVDIfsImage);
1337	AssertPtrReturn(pImage->pIfIo, VERR_INVALID_PARAMETER);
1338
1339	/* Create image file. */
1340	fOpen = VDOpenFlagsToFileOpenFlags(pImage->uOpenFlags, true /* fCreate */);
1341	rc = vdIfIoIntFileOpen(pImage->pIfIo, pImage->pszFilename, fOpen, &pImage->pStorage);
1342	if (RT_FAILURE(rc))
1343	{
1344	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: cannot create image '%s'"), pImage->pszFilename);
1345	goto out;
1346	}
1347
1348	/* Init image state. */
1349	pImage->cbSize = cbSize;
1350	pImage->cbCluster = QED_CLUSTER_SIZE_DEFAULT;
1351	pImage->cbTable = qedCluster2Byte(pImage, QED_TABLE_SIZE_DEFAULT);
1352	pImage->cTableEntries = pImage->cbTable / sizeof(uint64_t);
1353	pImage->offL1Table = qedCluster2Byte(pImage, 1); /* Cluster 0 is the header. */
1354	pImage->cbImage = (1 * pImage->cbCluster) + pImage->cbTable; /* Header + L1 table size. */
1355	pImage->cbBackingFilename = 0;
1356	pImage->offBackingFilename = 0;
1357	qedTableMasksInit(pImage);
1358
1359	/* Init L1 table. */
1360	pImage->paL1Table = (uint64_t *)RTMemAllocZ(pImage->cbTable);
1361	if (!pImage->paL1Table)
1362	{
1363	rc = vdIfError(pImage->pIfError, VERR_NO_MEMORY, RT_SRC_POS, N_("Qed: cannot allocate memory for L1 table of image '%s'"),
1364	pImage->pszFilename);
1365	goto out;
1366	}
1367
1368	rc = qedL2TblCacheCreate(pImage);
1369	if (RT_FAILURE(rc))
1370	{
1371	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Failed to create L2 cache for image '%s'"),
1372	pImage->pszFilename);
1373	goto out;
1374	}
1375
1376	if (RT_SUCCESS(rc) && pfnProgress)
1377	pfnProgress(pvUser, uPercentStart + uPercentSpan * 98 / 100);
1378
1379	rc = qedFlushImage(pImage);
1380
1381	out:
1382	if (RT_SUCCESS(rc) && pfnProgress)
1383	pfnProgress(pvUser, uPercentStart + uPercentSpan);
1384
1385	if (RT_FAILURE(rc))
1386	qedFreeImage(pImage, rc != VERR_ALREADY_EXISTS);
1387	return rc;
1388	}
1389
1390	/**
1391	* Rollback anything done during async cluster allocation.
1392	*
1393	* @returns VBox status code.
1394	* @param pImage The image instance data.
1395	* @param pIoCtx The I/O context.
1396	* @param pClusterAlloc The cluster allocation to rollback.
1397	*/
1398	static int qedAsyncClusterAllocRollback(PQEDIMAGE pImage, PVDIOCTX pIoCtx, PQEDCLUSTERASYNCALLOC pClusterAlloc)
1399	{
1400	RT_NOREF1(pIoCtx);
1401	int rc = VINF_SUCCESS;
1402
1403	switch (pClusterAlloc->enmAllocState)
1404	{
1405	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1406	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1407	{
1408	/* Assumption right now is that the L1 table is not modified if the link fails. */
1409	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1410	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1411	qedL2TblCacheEntryFree(pImage, pClusterAlloc->pL2Entry); /* Free it, it is not in the cache yet. */
1412	break;
1413	}
1414	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1415	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1416	{
1417	/* Assumption right now is that the L2 table is not modified if the link fails. */
1418	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage, pClusterAlloc->cbImageOld);
1419	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry); /* Release L2 cache entry. */
1420	break;
1421	}
1422	default:
1423	AssertMsgFailed(("Invalid cluster allocation state %d\n", pClusterAlloc->enmAllocState));
1424	rc = VERR_INVALID_STATE;
1425	}
1426
1427	RTMemFree(pClusterAlloc);
1428	return rc;
1429	}
1430
1431	/**
1432	* Updates the state of the async cluster allocation.
1433	*
1434	* @returns VBox status code.
1435	* @param pBackendData The opaque backend data.
1436	* @param pIoCtx I/O context associated with this request.
1437	* @param pvUser Opaque user data passed during a read/write request.
1438	* @param rcReq Status code for the completed request.
1439	*/
1440	static DECLCALLBACK(int) qedAsyncClusterAllocUpdate(void pBackendData, PVDIOCTX pIoCtx, void pvUser, int rcReq)
1441	{
1442	int rc = VINF_SUCCESS;
1443	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1444	PQEDCLUSTERASYNCALLOC pClusterAlloc = (PQEDCLUSTERASYNCALLOC)pvUser;
1445
1446	if (RT_FAILURE(rcReq))
1447	return qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1448
1449	AssertPtr(pClusterAlloc->pL2Entry);
1450
1451	switch (pClusterAlloc->enmAllocState)
1452	{
1453	case QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC:
1454	{
1455	uint64_t offUpdateLe = RT_H2LE_U64(pClusterAlloc->pL2Entry->offL2Tbl);
1456
1457	/* Update the link in the on disk L1 table now. */
1458	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_LINK;
1459	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1460	pImage->offL1Table + pClusterAlloc->idxL1*sizeof(uint64_t),
1461	&offUpdateLe, sizeof(uint64_t), pIoCtx,
1462	qedAsyncClusterAllocUpdate, pClusterAlloc);
1463	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1464	break;
1465	else if (RT_FAILURE(rc))
1466	{
1467	/* Rollback. */
1468	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1469	break;
1470	}
1471	/* Success, fall through. */
1472	}
1473	case QEDCLUSTERASYNCALLOCSTATE_L2_LINK:
1474	{
1475	/* L2 link updated in L1 , save L2 entry in cache and allocate new user data cluster. */
1476	uint64_t offData = qedClusterAllocate(pImage, 1);
1477
1478	/* Update the link in the in memory L1 table now. */
1479	pImage->paL1Table[pClusterAlloc->idxL1] = pClusterAlloc->pL2Entry->offL2Tbl;
1480	qedL2TblCacheEntryInsert(pImage, pClusterAlloc->pL2Entry);
1481
1482	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1483	pClusterAlloc->cbImageOld = offData;
1484	pClusterAlloc->offClusterNew = offData;
1485
1486	/* Write data. */
1487	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1488	offData, pIoCtx, pClusterAlloc->cbToWrite,
1489	qedAsyncClusterAllocUpdate, pClusterAlloc);
1490	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1491	break;
1492	else if (RT_FAILURE(rc))
1493	{
1494	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1495	RTMemFree(pClusterAlloc);
1496	break;
1497	}
1498	}
1499	case QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC:
1500	{
1501	uint64_t offUpdateLe = RT_H2LE_U64(pClusterAlloc->offClusterNew);
1502
1503	pClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_LINK;
1504
1505	/* Link L2 table and update it. */
1506	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1507	pImage->paL1Table[pClusterAlloc->idxL1] + pClusterAlloc->idxL2*sizeof(uint64_t),
1508	&offUpdateLe, sizeof(uint64_t), pIoCtx,
1509	qedAsyncClusterAllocUpdate, pClusterAlloc);
1510	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1511	break;
1512	else if (RT_FAILURE(rc))
1513	{
1514	qedAsyncClusterAllocRollback(pImage, pIoCtx, pClusterAlloc);
1515	RTMemFree(pClusterAlloc);
1516	break;
1517	}
1518	}
1519	case QEDCLUSTERASYNCALLOCSTATE_USER_LINK:
1520	{
1521	/* Everything done without errors, signal completion. */
1522	pClusterAlloc->pL2Entry->paL2Tbl[pClusterAlloc->idxL2] = pClusterAlloc->offClusterNew;
1523	qedL2TblCacheEntryRelease(pClusterAlloc->pL2Entry);
1524	RTMemFree(pClusterAlloc);
1525	rc = VINF_SUCCESS;
1526	break;
1527	}
1528	default:
1529	AssertMsgFailed(("Invalid async cluster allocation state %d\n",
1530	pClusterAlloc->enmAllocState));
1531	}
1532
1533	return rc;
1534	}
1535
1536	/** @copydoc VBOXHDDBACKEND::pfnCheckIfValid */
1537	static DECLCALLBACK(int) qedCheckIfValid(const char *pszFilename, PVDINTERFACE pVDIfsDisk,
1538	PVDINTERFACE pVDIfsImage, VDTYPE *penmType)
1539	{
1540	RT_NOREF1(pVDIfsDisk);
1541	LogFlowFunc(("pszFilename=\"%s\" pVDIfsDisk=%#p pVDIfsImage=%#p\n", pszFilename, pVDIfsDisk, pVDIfsImage));
1542	PVDIOSTORAGE pStorage = NULL;
1543	uint64_t cbFile;
1544	int rc = VINF_SUCCESS;
1545
1546	/* Get I/O interface. */
1547	PVDINTERFACEIOINT pIfIo = VDIfIoIntGet(pVDIfsImage);
1548	AssertPtrReturn(pIfIo, VERR_INVALID_PARAMETER);
1549
1550	if ( !VALID_PTR(pszFilename)
1551	\|\| !*pszFilename)
1552	{
1553	rc = VERR_INVALID_PARAMETER;
1554	goto out;
1555	}
1556
1557	/*
1558	* Open the file and read the footer.
1559	*/
1560	rc = vdIfIoIntFileOpen(pIfIo, pszFilename,
1561	VDOpenFlagsToFileOpenFlags(VD_OPEN_FLAGS_READONLY,
1562	false /* fCreate */),
1563	&pStorage);
1564	if (RT_SUCCESS(rc))
1565	{
1566	rc = vdIfIoIntFileGetSize(pIfIo, pStorage, &cbFile);
1567	if ( RT_SUCCESS(rc)
1568	&& cbFile > sizeof(QedHeader))
1569	{
1570	QedHeader Header;
1571
1572	rc = vdIfIoIntFileReadSync(pIfIo, pStorage, 0, &Header, sizeof(Header));
1573	if ( RT_SUCCESS(rc)
1574	&& qedHdrConvertToHostEndianess(&Header))
1575	{
1576	*penmType = VDTYPE_HDD;
1577	rc = VINF_SUCCESS;
1578	}
1579	else
1580	rc = VERR_VD_GEN_INVALID_HEADER;
1581	}
1582	else
1583	rc = VERR_VD_GEN_INVALID_HEADER;
1584	}
1585
1586	if (pStorage)
1587	vdIfIoIntFileClose(pIfIo, pStorage);
1588
1589	out:
1590	LogFlowFunc(("returns %Rrc\n", rc));
1591	return rc;
1592	}
1593
1594	/** @copydoc VBOXHDDBACKEND::pfnOpen */
1595	static DECLCALLBACK(int) qedOpen(const char *pszFilename, unsigned uOpenFlags,
1596	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1597	VDTYPE enmType, void **ppBackendData)
1598	{
1599	LogFlowFunc(("pszFilename=\"%s\" uOpenFlags=%#x pVDIfsDisk=%#p pVDIfsImage=%#p enmType=%u ppBackendData=%#p\n", pszFilename, uOpenFlags, pVDIfsDisk, pVDIfsImage, enmType, ppBackendData));
1600	int rc;
1601	PQEDIMAGE pImage;
1602
1603	NOREF(enmType); /*< @todo r=klaus make use of the type info. /
1604
1605	/* Check open flags. All valid flags are supported. */
1606	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1607	{
1608	rc = VERR_INVALID_PARAMETER;
1609	goto out;
1610	}
1611
1612	/* Check remaining arguments. */
1613	if ( !VALID_PTR(pszFilename)
1614	\|\| !*pszFilename)
1615	{
1616	rc = VERR_INVALID_PARAMETER;
1617	goto out;
1618	}
1619
1620
1621	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1622	if (!pImage)
1623	{
1624	rc = VERR_NO_MEMORY;
1625	goto out;
1626	}
1627	pImage->pszFilename = pszFilename;
1628	pImage->pStorage = NULL;
1629	pImage->pVDIfsDisk = pVDIfsDisk;
1630	pImage->pVDIfsImage = pVDIfsImage;
1631
1632	rc = qedOpenImage(pImage, uOpenFlags);
1633	if (RT_SUCCESS(rc))
1634	*ppBackendData = pImage;
1635	else
1636	RTMemFree(pImage);
1637
1638	out:
1639	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1640	return rc;
1641	}
1642
1643	/** @copydoc VBOXHDDBACKEND::pfnCreate */
1644	static DECLCALLBACK(int) qedCreate(const char *pszFilename, uint64_t cbSize,
1645	unsigned uImageFlags, const char *pszComment,
1646	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
1647	PCRTUUID pUuid, unsigned uOpenFlags,
1648	unsigned uPercentStart, unsigned uPercentSpan,
1649	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
1650	PVDINTERFACE pVDIfsOperation, VDTYPE enmType,
1651	void **ppBackendData)
1652	{
1653	RT_NOREF1(pUuid);
1654	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",
1655	pszFilename, cbSize, uImageFlags, pszComment, pPCHSGeometry, pLCHSGeometry, pUuid, uOpenFlags, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation, enmType, ppBackendData));
1656	int rc;
1657	PQEDIMAGE pImage;
1658
1659	PFNVDPROGRESS pfnProgress = NULL;
1660	void *pvUser = NULL;
1661	PVDINTERFACEPROGRESS pIfProgress = VDIfProgressGet(pVDIfsOperation);
1662	if (pIfProgress)
1663	{
1664	pfnProgress = pIfProgress->pfnProgress;
1665	pvUser = pIfProgress->Core.pvUser;
1666	}
1667
1668	/* Check the VD container type. */
1669	if (enmType != VDTYPE_HDD)
1670	{
1671	rc = VERR_VD_INVALID_TYPE;
1672	goto out;
1673	}
1674
1675	/* Check open flags. All valid flags are supported. */
1676	if (uOpenFlags & ~VD_OPEN_FLAGS_MASK)
1677	{
1678	rc = VERR_INVALID_PARAMETER;
1679	goto out;
1680	}
1681
1682	/* Check remaining arguments. */
1683	if ( !VALID_PTR(pszFilename)
1684	\|\| !*pszFilename
1685	\|\| !VALID_PTR(pPCHSGeometry)
1686	\|\| !VALID_PTR(pLCHSGeometry))
1687	{
1688	rc = VERR_INVALID_PARAMETER;
1689	goto out;
1690	}
1691
1692	pImage = (PQEDIMAGE)RTMemAllocZ(sizeof(QEDIMAGE));
1693	if (!pImage)
1694	{
1695	rc = VERR_NO_MEMORY;
1696	goto out;
1697	}
1698	pImage->pszFilename = pszFilename;
1699	pImage->pStorage = NULL;
1700	pImage->pVDIfsDisk = pVDIfsDisk;
1701	pImage->pVDIfsImage = pVDIfsImage;
1702
1703	rc = qedCreateImage(pImage, cbSize, uImageFlags, pszComment,
1704	pPCHSGeometry, pLCHSGeometry, uOpenFlags,
1705	pfnProgress, pvUser, uPercentStart, uPercentSpan);
1706	if (RT_SUCCESS(rc))
1707	{
1708	/* So far the image is opened in read/write mode. Make sure the
1709	* image is opened in read-only mode if the caller requested that. */
1710	if (uOpenFlags & VD_OPEN_FLAGS_READONLY)
1711	{
1712	qedFreeImage(pImage, false);
1713	rc = qedOpenImage(pImage, uOpenFlags);
1714	if (RT_FAILURE(rc))
1715	{
1716	RTMemFree(pImage);
1717	goto out;
1718	}
1719	}
1720	*ppBackendData = pImage;
1721	}
1722	else
1723	RTMemFree(pImage);
1724
1725	out:
1726	LogFlowFunc(("returns %Rrc (pBackendData=%#p)\n", rc, *ppBackendData));
1727	return rc;
1728	}
1729
1730	/** @copydoc VBOXHDDBACKEND::pfnRename */
1731	static DECLCALLBACK(int) qedRename(void pBackendData, const char pszFilename)
1732	{
1733	LogFlowFunc(("pBackendData=%#p pszFilename=%#p\n", pBackendData, pszFilename));
1734	int rc = VINF_SUCCESS;
1735	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1736
1737	/* Check arguments. */
1738	if ( !pImage
1739	\|\| !pszFilename
1740	\|\| !*pszFilename)
1741	{
1742	rc = VERR_INVALID_PARAMETER;
1743	goto out;
1744	}
1745
1746	/* Close the image. */
1747	rc = qedFreeImage(pImage, false);
1748	if (RT_FAILURE(rc))
1749	goto out;
1750
1751	/* Rename the file. */
1752	rc = vdIfIoIntFileMove(pImage->pIfIo, pImage->pszFilename, pszFilename, 0);
1753	if (RT_FAILURE(rc))
1754	{
1755	/* The move failed, try to reopen the original image. */
1756	int rc2 = qedOpenImage(pImage, pImage->uOpenFlags);
1757	if (RT_FAILURE(rc2))
1758	rc = rc2;
1759
1760	goto out;
1761	}
1762
1763	/* Update pImage with the new information. */
1764	pImage->pszFilename = pszFilename;
1765
1766	/* Open the old image with new name. */
1767	rc = qedOpenImage(pImage, pImage->uOpenFlags);
1768	if (RT_FAILURE(rc))
1769	goto out;
1770
1771	out:
1772	LogFlowFunc(("returns %Rrc\n", rc));
1773	return rc;
1774	}
1775
1776	/** @copydoc VBOXHDDBACKEND::pfnClose */
1777	static DECLCALLBACK(int) qedClose(void *pBackendData, bool fDelete)
1778	{
1779	LogFlowFunc(("pBackendData=%#p fDelete=%d\n", pBackendData, fDelete));
1780	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1781	int rc;
1782
1783	rc = qedFreeImage(pImage, fDelete);
1784	RTMemFree(pImage);
1785
1786	LogFlowFunc(("returns %Rrc\n", rc));
1787	return rc;
1788	}
1789
1790	static DECLCALLBACK(int) qedRead(void *pBackendData, uint64_t uOffset, size_t cbToRead,
1791	PVDIOCTX pIoCtx, size_t *pcbActuallyRead)
1792	{
1793	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToRead=%zu pcbActuallyRead=%#p\n",
1794	pBackendData, uOffset, pIoCtx, cbToRead, pcbActuallyRead));
1795	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1796	uint32_t offCluster = 0;
1797	uint32_t idxL1 = 0;
1798	uint32_t idxL2 = 0;
1799	uint64_t offFile = 0;
1800	int rc;
1801
1802	AssertPtr(pImage);
1803	Assert(uOffset % 512 == 0);
1804	Assert(cbToRead % 512 == 0);
1805
1806	if (!VALID_PTR(pIoCtx) \|\| !cbToRead)
1807	{
1808	rc = VERR_INVALID_PARAMETER;
1809	goto out;
1810	}
1811
1812	if ( uOffset + cbToRead > pImage->cbSize
1813	\|\| cbToRead == 0)
1814	{
1815	rc = VERR_INVALID_PARAMETER;
1816	goto out;
1817	}
1818
1819	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1820
1821	/* Clip read size to remain in the cluster. */
1822	cbToRead = RT_MIN(cbToRead, pImage->cbCluster - offCluster);
1823
1824	/* Get offset in image. */
1825	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offFile);
1826	if (RT_SUCCESS(rc))
1827	rc = vdIfIoIntFileReadUser(pImage->pIfIo, pImage->pStorage, offFile,
1828	pIoCtx, cbToRead);
1829
1830	if ( ( RT_SUCCESS(rc)
1831	\|\| rc == VERR_VD_BLOCK_FREE
1832	\|\| rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1833	&& pcbActuallyRead)
1834	*pcbActuallyRead = cbToRead;
1835
1836	out:
1837	LogFlowFunc(("returns %Rrc\n", rc));
1838	return rc;
1839	}
1840
1841	static DECLCALLBACK(int) qedWrite(void *pBackendData, uint64_t uOffset, size_t cbToWrite,
1842	PVDIOCTX pIoCtx, size_t pcbWriteProcess, size_t pcbPreRead,
1843	size_t *pcbPostRead, unsigned fWrite)
1844	{
1845	LogFlowFunc(("pBackendData=%#p uOffset=%llu pIoCtx=%#p cbToWrite=%zu pcbWriteProcess=%#p pcbPreRead=%#p pcbPostRead=%#p\n",
1846	pBackendData, uOffset, pIoCtx, cbToWrite, pcbWriteProcess, pcbPreRead, pcbPostRead));
1847	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
1848	uint32_t offCluster = 0;
1849	uint32_t idxL1 = 0;
1850	uint32_t idxL2 = 0;
1851	uint64_t offImage = 0;
1852	int rc = VINF_SUCCESS;
1853
1854	AssertPtr(pImage);
1855	Assert(!(uOffset % 512));
1856	Assert(!(cbToWrite % 512));
1857
1858	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
1859	{
1860	rc = VERR_VD_IMAGE_READ_ONLY;
1861	goto out;
1862	}
1863
1864	if (!VALID_PTR(pIoCtx) \|\| !cbToWrite)
1865	{
1866	rc = VERR_INVALID_PARAMETER;
1867	goto out;
1868	}
1869
1870	if ( uOffset + cbToWrite > pImage->cbSize
1871	\|\| cbToWrite == 0)
1872	{
1873	rc = VERR_INVALID_PARAMETER;
1874	goto out;
1875	}
1876
1877	/* Convert offset to L1, L2 index and cluster offset. */
1878	qedConvertLogicalOffset(pImage, uOffset, &idxL1, &idxL2, &offCluster);
1879
1880	/* Clip write size to remain in the cluster. */
1881	cbToWrite = RT_MIN(cbToWrite, pImage->cbCluster - offCluster);
1882	Assert(!(cbToWrite % 512));
1883
1884	/* Get offset in image. */
1885	rc = qedConvertToImageOffset(pImage, pIoCtx, idxL1, idxL2, offCluster, &offImage);
1886	if (RT_SUCCESS(rc))
1887	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1888	offImage, pIoCtx, cbToWrite, NULL, NULL);
1889	else if (rc == VERR_VD_BLOCK_FREE)
1890	{
1891	if ( cbToWrite == pImage->cbCluster
1892	&& !(fWrite & VD_WRITE_NO_ALLOC))
1893	{
1894	PQEDL2CACHEENTRY pL2Entry = NULL;
1895
1896	/* Full cluster write to previously unallocated cluster.
1897	* Allocate cluster and write data. */
1898	Assert(!offCluster);
1899
1900	do
1901	{
1902	/* Check if we have to allocate a new cluster for L2 tables. */
1903	if (!pImage->paL1Table[idxL1])
1904	{
1905	uint64_t offL2Tbl;
1906	PQEDCLUSTERASYNCALLOC pL2ClusterAlloc = NULL;
1907
1908	/* Allocate new async cluster allocation state. */
1909	pL2ClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1910	if (RT_UNLIKELY(!pL2ClusterAlloc))
1911	{
1912	rc = VERR_NO_MEMORY;
1913	break;
1914	}
1915
1916	pL2Entry = qedL2TblCacheEntryAlloc(pImage);
1917	if (!pL2Entry)
1918	{
1919	rc = VERR_NO_MEMORY;
1920	RTMemFree(pL2ClusterAlloc);
1921	break;
1922	}
1923
1924	offL2Tbl = qedClusterAllocate(pImage, qedByte2Cluster(pImage, pImage->cbTable));
1925	pL2Entry->offL2Tbl = offL2Tbl;
1926	memset(pL2Entry->paL2Tbl, 0, pImage->cbTable);
1927
1928	pL2ClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_L2_ALLOC;
1929	pL2ClusterAlloc->cbImageOld = offL2Tbl;
1930	pL2ClusterAlloc->offClusterNew = offL2Tbl;
1931	pL2ClusterAlloc->idxL1 = idxL1;
1932	pL2ClusterAlloc->idxL2 = idxL2;
1933	pL2ClusterAlloc->cbToWrite = cbToWrite;
1934	pL2ClusterAlloc->pL2Entry = pL2Entry;
1935
1936	/*
1937	* Write the L2 table first and link to the L1 table afterwards.
1938	* If something unexpected happens the worst case which can happen
1939	* is a leak of some clusters.
1940	*/
1941	rc = vdIfIoIntFileWriteMeta(pImage->pIfIo, pImage->pStorage,
1942	offL2Tbl, pL2Entry->paL2Tbl, pImage->cbTable, pIoCtx,
1943	qedAsyncClusterAllocUpdate, pL2ClusterAlloc);
1944	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1945	break;
1946	else if (RT_FAILURE(rc))
1947	{
1948	RTMemFree(pL2ClusterAlloc);
1949	qedL2TblCacheEntryFree(pImage, pL2Entry);
1950	break;
1951	}
1952
1953	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pL2ClusterAlloc, rc);
1954	}
1955	else
1956	{
1957	rc = qedL2TblCacheFetchAsync(pImage, pIoCtx, pImage->paL1Table[idxL1],
1958	&pL2Entry);
1959
1960	if (RT_SUCCESS(rc))
1961	{
1962	PQEDCLUSTERASYNCALLOC pDataClusterAlloc = NULL;
1963
1964	/* Allocate new async cluster allocation state. */
1965	pDataClusterAlloc = (PQEDCLUSTERASYNCALLOC)RTMemAllocZ(sizeof(QEDCLUSTERASYNCALLOC));
1966	if (RT_UNLIKELY(!pDataClusterAlloc))
1967	{
1968	rc = VERR_NO_MEMORY;
1969	break;
1970	}
1971
1972	/* Allocate new cluster for the data. */
1973	uint64_t offData = qedClusterAllocate(pImage, 1);
1974
1975	pDataClusterAlloc->enmAllocState = QEDCLUSTERASYNCALLOCSTATE_USER_ALLOC;
1976	pDataClusterAlloc->cbImageOld = offData;
1977	pDataClusterAlloc->offClusterNew = offData;
1978	pDataClusterAlloc->idxL1 = idxL1;
1979	pDataClusterAlloc->idxL2 = idxL2;
1980	pDataClusterAlloc->cbToWrite = cbToWrite;
1981	pDataClusterAlloc->pL2Entry = pL2Entry;
1982
1983	/* Write data. */
1984	rc = vdIfIoIntFileWriteUser(pImage->pIfIo, pImage->pStorage,
1985	offData, pIoCtx, cbToWrite,
1986	qedAsyncClusterAllocUpdate, pDataClusterAlloc);
1987	if (rc == VERR_VD_ASYNC_IO_IN_PROGRESS)
1988	break;
1989	else if (RT_FAILURE(rc))
1990	{
1991	RTMemFree(pDataClusterAlloc);
1992	break;
1993	}
1994
1995	rc = qedAsyncClusterAllocUpdate(pImage, pIoCtx, pDataClusterAlloc, rc);
1996	}
1997	}
1998
1999	} while (0);
2000
2001	*pcbPreRead = 0;
2002	*pcbPostRead = 0;
2003	}
2004	else
2005	{
2006	/* Trying to do a partial write to an unallocated cluster. Don't do
2007	* anything except letting the upper layer know what to do. */
2008	*pcbPreRead = offCluster;
2009	pcbPostRead = pImage->cbCluster - cbToWrite - pcbPreRead;
2010	}
2011	}
2012
2013	if (pcbWriteProcess)
2014	*pcbWriteProcess = cbToWrite;
2015
2016
2017	out:
2018	LogFlowFunc(("returns %Rrc\n", rc));
2019	return rc;
2020	}
2021
2022	static DECLCALLBACK(int) qedFlush(void *pBackendData, PVDIOCTX pIoCtx)
2023	{
2024	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2025	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2026	int rc = VINF_SUCCESS;
2027
2028	Assert(pImage);
2029
2030	if (VALID_PTR(pIoCtx))
2031	rc = qedFlushImageAsync(pImage, pIoCtx);
2032	else
2033	rc = VERR_INVALID_PARAMETER;
2034
2035	LogFlowFunc(("returns %Rrc\n", rc));
2036	return rc;
2037	}
2038
2039	/** @copydoc VBOXHDDBACKEND::pfnGetVersion */
2040	static DECLCALLBACK(unsigned) qedGetVersion(void *pBackendData)
2041	{
2042	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2043	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2044
2045	AssertPtr(pImage);
2046
2047	if (pImage)
2048	return 1;
2049	else
2050	return 0;
2051	}
2052
2053	/** @copydoc VBOXHDDBACKEND::pfnGetSectorSize */
2054	static DECLCALLBACK(uint32_t) qedGetSectorSize(void *pBackendData)
2055	{
2056	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2057	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2058	uint32_t cb = 0;
2059
2060	AssertPtr(pImage);
2061
2062	if (pImage && pImage->pStorage)
2063	cb = 512;
2064
2065	LogFlowFunc(("returns %u\n", cb));
2066	return cb;
2067	}
2068
2069	/** @copydoc VBOXHDDBACKEND::pfnGetSize */
2070	static DECLCALLBACK(uint64_t) qedGetSize(void *pBackendData)
2071	{
2072	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2073	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2074	uint64_t cb = 0;
2075
2076	AssertPtr(pImage);
2077
2078	if (pImage && pImage->pStorage)
2079	cb = pImage->cbSize;
2080
2081	LogFlowFunc(("returns %llu\n", cb));
2082	return cb;
2083	}
2084
2085	/** @copydoc VBOXHDDBACKEND::pfnGetFileSize */
2086	static DECLCALLBACK(uint64_t) qedGetFileSize(void *pBackendData)
2087	{
2088	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2089	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2090	uint64_t cb = 0;
2091
2092	AssertPtr(pImage);
2093
2094	if (pImage)
2095	{
2096	uint64_t cbFile;
2097	if (pImage->pStorage)
2098	{
2099	int rc = vdIfIoIntFileGetSize(pImage->pIfIo, pImage->pStorage, &cbFile);
2100	if (RT_SUCCESS(rc))
2101	cb += cbFile;
2102	}
2103	}
2104
2105	LogFlowFunc(("returns %lld\n", cb));
2106	return cb;
2107	}
2108
2109	/** @copydoc VBOXHDDBACKEND::pfnGetPCHSGeometry */
2110	static DECLCALLBACK(int) qedGetPCHSGeometry(void *pBackendData,
2111	PVDGEOMETRY pPCHSGeometry)
2112	{
2113	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p\n", pBackendData, pPCHSGeometry));
2114	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2115	int rc;
2116
2117	AssertPtr(pImage);
2118
2119	if (pImage)
2120	{
2121	if (pImage->PCHSGeometry.cCylinders)
2122	{
2123	*pPCHSGeometry = pImage->PCHSGeometry;
2124	rc = VINF_SUCCESS;
2125	}
2126	else
2127	rc = VERR_VD_GEOMETRY_NOT_SET;
2128	}
2129	else
2130	rc = VERR_VD_NOT_OPENED;
2131
2132	LogFlowFunc(("returns %Rrc (PCHS=%u/%u/%u)\n", rc, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2133	return rc;
2134	}
2135
2136	/** @copydoc VBOXHDDBACKEND::pfnSetPCHSGeometry */
2137	static DECLCALLBACK(int) qedSetPCHSGeometry(void *pBackendData,
2138	PCVDGEOMETRY pPCHSGeometry)
2139	{
2140	LogFlowFunc(("pBackendData=%#p pPCHSGeometry=%#p PCHS=%u/%u/%u\n", pBackendData, pPCHSGeometry, pPCHSGeometry->cCylinders, pPCHSGeometry->cHeads, pPCHSGeometry->cSectors));
2141	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2142	int rc;
2143
2144	AssertPtr(pImage);
2145
2146	if (pImage)
2147	{
2148	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2149	{
2150	rc = VERR_VD_IMAGE_READ_ONLY;
2151	goto out;
2152	}
2153
2154	pImage->PCHSGeometry = *pPCHSGeometry;
2155	rc = VINF_SUCCESS;
2156	}
2157	else
2158	rc = VERR_VD_NOT_OPENED;
2159
2160	out:
2161	LogFlowFunc(("returns %Rrc\n", rc));
2162	return rc;
2163	}
2164
2165	/** @copydoc VBOXHDDBACKEND::pfnGetLCHSGeometry */
2166	static DECLCALLBACK(int) qedGetLCHSGeometry(void *pBackendData,
2167	PVDGEOMETRY pLCHSGeometry)
2168	{
2169	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p\n", pBackendData, pLCHSGeometry));
2170	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2171	int rc;
2172
2173	AssertPtr(pImage);
2174
2175	if (pImage)
2176	{
2177	if (pImage->LCHSGeometry.cCylinders)
2178	{
2179	*pLCHSGeometry = pImage->LCHSGeometry;
2180	rc = VINF_SUCCESS;
2181	}
2182	else
2183	rc = VERR_VD_GEOMETRY_NOT_SET;
2184	}
2185	else
2186	rc = VERR_VD_NOT_OPENED;
2187
2188	LogFlowFunc(("returns %Rrc (LCHS=%u/%u/%u)\n", rc, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2189	return rc;
2190	}
2191
2192	/** @copydoc VBOXHDDBACKEND::pfnSetLCHSGeometry */
2193	static DECLCALLBACK(int) qedSetLCHSGeometry(void *pBackendData,
2194	PCVDGEOMETRY pLCHSGeometry)
2195	{
2196	LogFlowFunc(("pBackendData=%#p pLCHSGeometry=%#p LCHS=%u/%u/%u\n", pBackendData, pLCHSGeometry, pLCHSGeometry->cCylinders, pLCHSGeometry->cHeads, pLCHSGeometry->cSectors));
2197	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2198	int rc;
2199
2200	AssertPtr(pImage);
2201
2202	if (pImage)
2203	{
2204	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2205	{
2206	rc = VERR_VD_IMAGE_READ_ONLY;
2207	goto out;
2208	}
2209
2210	pImage->LCHSGeometry = *pLCHSGeometry;
2211	rc = VINF_SUCCESS;
2212	}
2213	else
2214	rc = VERR_VD_NOT_OPENED;
2215
2216	out:
2217	LogFlowFunc(("returns %Rrc\n", rc));
2218	return rc;
2219	}
2220
2221	/** @copydoc VBOXHDDBACKEND::pfnGetImageFlags */
2222	static DECLCALLBACK(unsigned) qedGetImageFlags(void *pBackendData)
2223	{
2224	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2225	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2226	unsigned uImageFlags;
2227
2228	AssertPtr(pImage);
2229
2230	if (pImage)
2231	uImageFlags = pImage->uImageFlags;
2232	else
2233	uImageFlags = 0;
2234
2235	LogFlowFunc(("returns %#x\n", uImageFlags));
2236	return uImageFlags;
2237	}
2238
2239	/** @copydoc VBOXHDDBACKEND::pfnGetOpenFlags */
2240	static DECLCALLBACK(unsigned) qedGetOpenFlags(void *pBackendData)
2241	{
2242	LogFlowFunc(("pBackendData=%#p\n", pBackendData));
2243	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2244	unsigned uOpenFlags;
2245
2246	AssertPtr(pImage);
2247
2248	if (pImage)
2249	uOpenFlags = pImage->uOpenFlags;
2250	else
2251	uOpenFlags = 0;
2252
2253	LogFlowFunc(("returns %#x\n", uOpenFlags));
2254	return uOpenFlags;
2255	}
2256
2257	/** @copydoc VBOXHDDBACKEND::pfnSetOpenFlags */
2258	static DECLCALLBACK(int) qedSetOpenFlags(void *pBackendData, unsigned uOpenFlags)
2259	{
2260	LogFlowFunc(("pBackendData=%#p\n uOpenFlags=%#x", pBackendData, uOpenFlags));
2261	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2262	int rc;
2263
2264	/* Image must be opened and the new flags must be valid. */
2265	if (!pImage \|\| (uOpenFlags & ~( VD_OPEN_FLAGS_READONLY \| VD_OPEN_FLAGS_INFO
2266	\| VD_OPEN_FLAGS_ASYNC_IO \| VD_OPEN_FLAGS_SKIP_CONSISTENCY_CHECKS)))
2267	{
2268	rc = VERR_INVALID_PARAMETER;
2269	goto out;
2270	}
2271
2272	/* Implement this operation via reopening the image. */
2273	rc = qedFreeImage(pImage, false);
2274	if (RT_FAILURE(rc))
2275	goto out;
2276	rc = qedOpenImage(pImage, uOpenFlags);
2277
2278	out:
2279	LogFlowFunc(("returns %Rrc\n", rc));
2280	return rc;
2281	}
2282
2283	/** @copydoc VBOXHDDBACKEND::pfnGetComment */
2284	static DECLCALLBACK(int) qedGetComment(void pBackendData, char pszComment,
2285	size_t cbComment)
2286	{
2287	RT_NOREF2(pszComment, cbComment);
2288	LogFlowFunc(("pBackendData=%#p pszComment=%#p cbComment=%zu\n", pBackendData, pszComment, cbComment));
2289	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2290	int rc;
2291
2292	AssertPtr(pImage);
2293
2294	if (pImage)
2295	rc = VERR_NOT_SUPPORTED;
2296	else
2297	rc = VERR_VD_NOT_OPENED;
2298
2299	LogFlowFunc(("returns %Rrc comment='%s'\n", rc, pszComment));
2300	return rc;
2301	}
2302
2303	/** @copydoc VBOXHDDBACKEND::pfnSetComment */
2304	static DECLCALLBACK(int) qedSetComment(void pBackendData, const char pszComment)
2305	{
2306	RT_NOREF1(pszComment);
2307	LogFlowFunc(("pBackendData=%#p pszComment=\"%s\"\n", pBackendData, pszComment));
2308	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2309	int rc;
2310
2311	AssertPtr(pImage);
2312
2313	if (pImage)
2314	{
2315	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2316	rc = VERR_VD_IMAGE_READ_ONLY;
2317	else
2318	rc = VERR_NOT_SUPPORTED;
2319	}
2320	else
2321	rc = VERR_VD_NOT_OPENED;
2322
2323	LogFlowFunc(("returns %Rrc\n", rc));
2324	return rc;
2325	}
2326
2327	/** @copydoc VBOXHDDBACKEND::pfnGetUuid */
2328	static DECLCALLBACK(int) qedGetUuid(void *pBackendData, PRTUUID pUuid)
2329	{
2330	RT_NOREF1(pUuid);
2331	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2332	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2333	int rc;
2334
2335	AssertPtr(pImage);
2336
2337	if (pImage)
2338	rc = VERR_NOT_SUPPORTED;
2339	else
2340	rc = VERR_VD_NOT_OPENED;
2341
2342	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2343	return rc;
2344	}
2345
2346	/** @copydoc VBOXHDDBACKEND::pfnSetUuid */
2347	static DECLCALLBACK(int) qedSetUuid(void *pBackendData, PCRTUUID pUuid)
2348	{
2349	RT_NOREF1(pUuid);
2350	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2351	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2352	int rc;
2353
2354	LogFlowFunc(("%RTuuid\n", pUuid));
2355	AssertPtr(pImage);
2356
2357	if (pImage)
2358	{
2359	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2360	rc = VERR_NOT_SUPPORTED;
2361	else
2362	rc = VERR_VD_IMAGE_READ_ONLY;
2363	}
2364	else
2365	rc = VERR_VD_NOT_OPENED;
2366
2367	LogFlowFunc(("returns %Rrc\n", rc));
2368	return rc;
2369	}
2370
2371	/** @copydoc VBOXHDDBACKEND::pfnGetModificationUuid */
2372	static DECLCALLBACK(int) qedGetModificationUuid(void *pBackendData, PRTUUID pUuid)
2373	{
2374	RT_NOREF1(pUuid);
2375	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2376	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2377	int rc;
2378
2379	AssertPtr(pImage);
2380
2381	if (pImage)
2382	rc = VERR_NOT_SUPPORTED;
2383	else
2384	rc = VERR_VD_NOT_OPENED;
2385
2386	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2387	return rc;
2388	}
2389
2390	/** @copydoc VBOXHDDBACKEND::pfnSetModificationUuid */
2391	static DECLCALLBACK(int) qedSetModificationUuid(void *pBackendData, PCRTUUID pUuid)
2392	{
2393	RT_NOREF1(pUuid);
2394	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2395	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2396	int rc;
2397
2398	AssertPtr(pImage);
2399
2400	if (pImage)
2401	{
2402	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2403	rc = VERR_NOT_SUPPORTED;
2404	else
2405	rc = VERR_VD_IMAGE_READ_ONLY;
2406	}
2407	else
2408	rc = VERR_VD_NOT_OPENED;
2409
2410	LogFlowFunc(("returns %Rrc\n", rc));
2411	return rc;
2412	}
2413
2414	/** @copydoc VBOXHDDBACKEND::pfnGetParentUuid */
2415	static DECLCALLBACK(int) qedGetParentUuid(void *pBackendData, PRTUUID pUuid)
2416	{
2417	RT_NOREF1(pUuid);
2418	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2419	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2420	int rc;
2421
2422	AssertPtr(pImage);
2423
2424	if (pImage)
2425	rc = VERR_NOT_SUPPORTED;
2426	else
2427	rc = VERR_VD_NOT_OPENED;
2428
2429	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2430	return rc;
2431	}
2432
2433	/** @copydoc VBOXHDDBACKEND::pfnSetParentUuid */
2434	static DECLCALLBACK(int) qedSetParentUuid(void *pBackendData, PCRTUUID pUuid)
2435	{
2436	RT_NOREF1(pUuid);
2437	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2438	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2439	int rc;
2440
2441	AssertPtr(pImage);
2442
2443	if (pImage)
2444	{
2445	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2446	rc = VERR_NOT_SUPPORTED;
2447	else
2448	rc = VERR_VD_IMAGE_READ_ONLY;
2449	}
2450	else
2451	rc = VERR_VD_NOT_OPENED;
2452
2453	LogFlowFunc(("returns %Rrc\n", rc));
2454	return rc;
2455	}
2456
2457	/** @copydoc VBOXHDDBACKEND::pfnGetParentModificationUuid */
2458	static DECLCALLBACK(int) qedGetParentModificationUuid(void *pBackendData, PRTUUID pUuid)
2459	{
2460	RT_NOREF1(pUuid);
2461	LogFlowFunc(("pBackendData=%#p pUuid=%#p\n", pBackendData, pUuid));
2462	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2463	int rc;
2464
2465	AssertPtr(pImage);
2466
2467	if (pImage)
2468	rc = VERR_NOT_SUPPORTED;
2469	else
2470	rc = VERR_VD_NOT_OPENED;
2471
2472	LogFlowFunc(("returns %Rrc (%RTuuid)\n", rc, pUuid));
2473	return rc;
2474	}
2475
2476	/** @copydoc VBOXHDDBACKEND::pfnSetParentModificationUuid */
2477	static DECLCALLBACK(int) qedSetParentModificationUuid(void *pBackendData, PCRTUUID pUuid)
2478	{
2479	RT_NOREF1(pUuid);
2480	LogFlowFunc(("pBackendData=%#p Uuid=%RTuuid\n", pBackendData, pUuid));
2481	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2482	int rc;
2483
2484	AssertPtr(pImage);
2485
2486	if (pImage)
2487	{
2488	if (!(pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY))
2489	rc = VERR_NOT_SUPPORTED;
2490	else
2491	rc = VERR_VD_IMAGE_READ_ONLY;
2492	}
2493	else
2494	rc = VERR_VD_NOT_OPENED;
2495
2496	LogFlowFunc(("returns %Rrc\n", rc));
2497	return rc;
2498	}
2499
2500	/** @copydoc VBOXHDDBACKEND::pfnDump */
2501	static DECLCALLBACK(void) qedDump(void *pBackendData)
2502	{
2503	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2504
2505	AssertPtr(pImage);
2506	if (pImage)
2507	{
2508	vdIfErrorMessage(pImage->pIfError, "Header: Geometry PCHS=%u/%u/%u LCHS=%u/%u/%u cbSector=%llu\n",
2509	pImage->PCHSGeometry.cCylinders, pImage->PCHSGeometry.cHeads, pImage->PCHSGeometry.cSectors,
2510	pImage->LCHSGeometry.cCylinders, pImage->LCHSGeometry.cHeads, pImage->LCHSGeometry.cSectors,
2511	pImage->cbSize / 512);
2512	}
2513	}
2514
2515	/** @copydoc VBOXHDDBACKEND::pfnGetParentFilename */
2516	static DECLCALLBACK(int) qedGetParentFilename(void pBackendData, char *ppszParentFilename)
2517	{
2518	int rc = VINF_SUCCESS;
2519	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2520
2521	AssertPtr(pImage);
2522	if (pImage)
2523	if (pImage->pszBackingFilename)
2524	*ppszParentFilename = RTStrDup(pImage->pszBackingFilename);
2525	else
2526	rc = VERR_NOT_SUPPORTED;
2527	else
2528	rc = VERR_VD_NOT_OPENED;
2529
2530	LogFlowFunc(("returns %Rrc\n", rc));
2531	return rc;
2532	}
2533
2534	/** @copydoc VBOXHDDBACKEND::pfnSetParentFilename */
2535	static DECLCALLBACK(int) qedSetParentFilename(void pBackendData, const char pszParentFilename)
2536	{
2537	int rc = VINF_SUCCESS;
2538	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2539
2540	AssertPtr(pImage);
2541	if (pImage)
2542	{
2543	if (pImage->uOpenFlags & VD_OPEN_FLAGS_READONLY)
2544	rc = VERR_VD_IMAGE_READ_ONLY;
2545	else if ( pImage->pszBackingFilename
2546	&& (strlen(pszParentFilename) > pImage->cbBackingFilename))
2547	rc = VERR_NOT_SUPPORTED; /* The new filename is longer than the old one. */
2548	else
2549	{
2550	if (pImage->pszBackingFilename)
2551	RTStrFree(pImage->pszBackingFilename);
2552	pImage->pszBackingFilename = RTStrDup(pszParentFilename);
2553	if (!pImage->pszBackingFilename)
2554	rc = VERR_NO_MEMORY;
2555	else
2556	{
2557	if (!pImage->offBackingFilename)
2558	{
2559	/* Allocate new cluster. */
2560	uint64_t offData = qedClusterAllocate(pImage, 1);
2561
2562	Assert((offData & UINT32_MAX) == offData);
2563	pImage->offBackingFilename = (uint32_t)offData;
2564	pImage->cbBackingFilename = (uint32_t)strlen(pszParentFilename);
2565	rc = vdIfIoIntFileSetSize(pImage->pIfIo, pImage->pStorage,
2566	offData + pImage->cbCluster);
2567	}
2568
2569	if (RT_SUCCESS(rc))
2570	rc = vdIfIoIntFileWriteSync(pImage->pIfIo, pImage->pStorage,
2571	pImage->offBackingFilename,
2572	pImage->pszBackingFilename,
2573	strlen(pImage->pszBackingFilename));
2574	}
2575	}
2576	}
2577	else
2578	rc = VERR_VD_NOT_OPENED;
2579
2580	LogFlowFunc(("returns %Rrc\n", rc));
2581	return rc;
2582	}
2583
2584	/** @copydoc VBOXHDDBACKEND::pfnResize */
2585	static DECLCALLBACK(int) qedResize(void *pBackendData, uint64_t cbSize,
2586	PCVDGEOMETRY pPCHSGeometry, PCVDGEOMETRY pLCHSGeometry,
2587	unsigned uPercentStart, unsigned uPercentSpan,
2588	PVDINTERFACE pVDIfsDisk, PVDINTERFACE pVDIfsImage,
2589	PVDINTERFACE pVDIfsOperation)
2590	{
2591	RT_NOREF7(pPCHSGeometry, pLCHSGeometry, uPercentStart, uPercentSpan, pVDIfsDisk, pVDIfsImage, pVDIfsOperation);
2592	PQEDIMAGE pImage = (PQEDIMAGE)pBackendData;
2593	int rc = VINF_SUCCESS;
2594
2595	/* Making the image smaller is not supported at the moment. */
2596	if (cbSize < pImage->cbSize)
2597	rc = VERR_NOT_SUPPORTED;
2598	else if (cbSize > pImage->cbSize)
2599	{
2600	/*
2601	* It is enough to just update the size field in the header to complete
2602	* growing. With the default cluster and table sizes the image can be expanded
2603	* to 64TB without overflowing the L1 and L2 tables making block relocation
2604	* superfluous.
2605	* @todo: The rare case where block relocation is still required (non default
2606	* table and/or cluster size or images with more than 64TB) is not
2607	* implemented yet and resizing such an image will fail with an error.
2608	*/
2609	if (qedByte2Cluster(pImage, pImage->cbTable)pImage->cTableEntriespImage->cTableEntries*pImage->cbCluster < cbSize)
2610	rc = vdIfError(pImage->pIfError, VERR_BUFFER_OVERFLOW, RT_SRC_POS,
2611	N_("Qed: Resizing the image '%s' is not supported because it would overflow the L1 and L2 table\n"),
2612	pImage->pszFilename);
2613	else
2614	{
2615	uint64_t cbSizeOld = pImage->cbSize;
2616
2617	pImage->cbSize = cbSize;
2618	rc = qedFlushImage(pImage);
2619	if (RT_FAILURE(rc))
2620	{
2621	pImage->cbSize = cbSizeOld; /* Restore */
2622
2623	rc = vdIfError(pImage->pIfError, rc, RT_SRC_POS, N_("Qed: Resizing the image '%s' failed\n"),
2624	pImage->pszFilename);
2625	}
2626	}
2627	}
2628	/* Same size doesn't change the image at all. */
2629
2630	LogFlowFunc(("returns %Rrc\n", rc));
2631	return rc;
2632	}
2633
2634
2635	const VBOXHDDBACKEND g_QedBackend =
2636	{
2637	/* pszBackendName */
2638	"QED",
2639	/* cbSize */
2640	sizeof(VBOXHDDBACKEND),
2641	/* uBackendCaps */
2642	VD_CAP_FILE \| VD_CAP_VFS \| VD_CAP_CREATE_DYNAMIC \| VD_CAP_DIFF \| VD_CAP_ASYNC,
2643	/* paFileExtensions */
2644	s_aQedFileExtensions,
2645	/* paConfigInfo */
2646	NULL,
2647	/* pfnCheckIfValid */
2648	qedCheckIfValid,
2649	/* pfnOpen */
2650	qedOpen,
2651	/* pfnCreate */
2652	qedCreate,
2653	/* pfnRename */
2654	qedRename,
2655	/* pfnClose */
2656	qedClose,
2657	/* pfnRead */
2658	qedRead,
2659	/* pfnWrite */
2660	qedWrite,
2661	/* pfnFlush */
2662	qedFlush,
2663	/* pfnDiscard */
2664	NULL,
2665	/* pfnGetVersion */
2666	qedGetVersion,
2667	/* pfnGetSectorSize */
2668	qedGetSectorSize,
2669	/* pfnGetSize */
2670	qedGetSize,
2671	/* pfnGetFileSize */
2672	qedGetFileSize,
2673	/* pfnGetPCHSGeometry */
2674	qedGetPCHSGeometry,
2675	/* pfnSetPCHSGeometry */
2676	qedSetPCHSGeometry,
2677	/* pfnGetLCHSGeometry */
2678	qedGetLCHSGeometry,
2679	/* pfnSetLCHSGeometry */
2680	qedSetLCHSGeometry,
2681	/* pfnGetImageFlags */
2682	qedGetImageFlags,
2683	/* pfnGetOpenFlags */
2684	qedGetOpenFlags,
2685	/* pfnSetOpenFlags */
2686	qedSetOpenFlags,
2687	/* pfnGetComment */
2688	qedGetComment,
2689	/* pfnSetComment */
2690	qedSetComment,
2691	/* pfnGetUuid */
2692	qedGetUuid,
2693	/* pfnSetUuid */
2694	qedSetUuid,
2695	/* pfnGetModificationUuid */
2696	qedGetModificationUuid,
2697	/* pfnSetModificationUuid */
2698	qedSetModificationUuid,
2699	/* pfnGetParentUuid */
2700	qedGetParentUuid,
2701	/* pfnSetParentUuid */
2702	qedSetParentUuid,
2703	/* pfnGetParentModificationUuid */
2704	qedGetParentModificationUuid,
2705	/* pfnSetParentModificationUuid */
2706	qedSetParentModificationUuid,
2707	/* pfnDump */
2708	qedDump,
2709	/* pfnGetTimestamp */
2710	NULL,
2711	/* pfnGetParentTimestamp */
2712	NULL,
2713	/* pfnSetParentTimestamp */
2714	NULL,
2715	/* pfnGetParentFilename */
2716	qedGetParentFilename,
2717	/* pfnSetParentFilename */
2718	qedSetParentFilename,
2719	/* pfnComposeLocation */
2720	genericFileComposeLocation,
2721	/* pfnComposeName */
2722	genericFileComposeName,
2723	/* pfnCompact */
2724	NULL,
2725	/* pfnResize */
2726	qedResize,
2727	/* pfnRepair */
2728	NULL,
2729	/* pfnTraverseMetadata */
2730	NULL
2731	};

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

source: vbox/trunk/src/VBox/Storage/QED.cpp@ 62746

Download in other formats: