DrvVD: don't make 'rc=%Rrc' part of the error message as the error code is printed later anyway

Add support to supply passwords for disk encryption while the VM is running

Storage/VD: Allow different filter chains for reads and writes (required by disk image encryption to encrypt images afterwards)

Storage/iSCSI: Fail when a connection attempt takes too long to pause the VM more quickly when the target is not responding

Disk encryption: Make sure the DekMissing guest property is set before the state change handler is called when the VM is suspended

Port a part of r95116 (disk encryption changes in 4.3) which was forgotten to trunk. Fixes starting a VM with an encrypted VHD image

build fix

build fix

pdmifs.h,DrvBlock,DrvVD: Add methods to allocate and free I/O buffer memory to PDMIBLOCK and PDMIMEDIA. DrvVD decides how to allocate the memory based on the setup

Console,DrvVD: Clear the encryption keys on suspend

Console,DrvVD: Make disk encryption completely configurable through medium properties

Main,DrvVD,VD: Implement support for VD plugins from extension packs

Storage,DrvVD,Main: Redo the way secret keys are passed from Main to encryption filters in DrvVD. Instead of using the config interface in VD use a distinct interface for retrieving secret keys which directly talks to the entity storing the keys in Console. Avoids copying sensitive data into probably insecure buffers. Key consumers only get a reference to the key which they can use and don't have to worry about allocating secure key memory.

Main,DrvVD: Interface to pass the keys to the disk encryption module from Main

Devics/Storage/DrvVD: Add support for filter plugins

Devices: Whitespace and svn:keyword cleanups by scm.

DrvVD.cpp: nits and some logging.

Devices/Storage: Support different sector sizes, booting from images with a sector size != 512 is not supported

Devices/Storage: Preparations for non 512 byte sector sizes. Add a new getter in the block and media interfaces and remove uses of 512 constants and use the reported sector size from the underlying medium

tabs

Review of PDM driver destructors making sure that variables they use are correctly initialized in the constructor. Found several RTFileClose(0) cases.

PDMAsyncCompletion: PVM -> PUVM (one instance), internalize internal APIs where possible (not all because of test cases). API docs lives where the implmentation lives, NOT in headers (IPRT is the exception).

DrvVD: IPv6 and new LWIP introduction.

Storage: Introduce new flag to skip unnecessary consistency checks in the VMDK backend while opening an image in readonly mode to speed up opening snapshots

Fix error message

DrvVD: Add InformAboutZeroBlocks CFGM key

*: gcc-4.7: ~0 => ~0U in initializers (''warning: narrowing conversion of `-1' from `int' to `unsigned int' inside { } is ill-formed in C++11 [-Wnarrowing]'')

Various drivers: Use PDMIBASE_2_PDMDRV instead of cooking your own stuff.

PDMBlkCache: Clear all entries on reset to prevent that unwritten data is in the cache which can lead to data corruption in case the cache isn't used after the reset

PDM,Devices,Drivers: Add async discard API and make us of it

AHCI+DrvBlock+DrvVD: Add support for the TRIM command and connect with the VD discard support.

VD: Interface cleanup. Merge the two involved structures (generic interface descriptor and callback table) into one, remove the duplicated interface wrappers in the backends and move the interface definitions into separate headers separating public and private interfaces.

DrvVD: Correct logging\n

DrvVD: Fix crash for the internal network stack case

DrvVD: Make use of the new flag in case UseNewIo is unset but there is a bandwidth group assigned

Move the misc files the in src/VBox/Devices/ directory into a build/ subdirectory, changing their names to match the target module.

VMM reorg: Moving the public include files from include/VBox to include/VBox/vmm.

BlkCache: Fix possible crash during suspend of the VM. The cache must be suspended before making the disk readonly to make sure all I/O completed

DrvVD: remove poweroff callback (not needed anymore). Fixes assertion during destruction if there is still dirty data in the block cache

DrvVD: Don't use the builtin cache when opening the image on readonly mode

Switch to the new block cache and disable the old one for now

Storage: Introduce interface to query the location of a medium (device + instance + LUN) in a VM to generate IDs which are constant for saved states

DrvVD: Bugfix

DrvVD: Block cache integration, work in progress

VD: Use S/G buffers instead of S/G arrays for VDAsyncWrite and VDAsyncRead (Preparation for moving the I/O cache in front of VD, see #5295)

VCI: Updates

VD: Move the generic virtual disk framework + backends to src/VBox/Storage and rename the files to get rid of the HDD part because it supports floppy and DVD images too

*: spelling fixes, thanks Timeless!

Storage: Implement offical support for other disk types like DVD and floppy images. DMG images can be used now without hacks

AsyncCompletion: Move bandwidth groups to the generic code and support arbitrary grouping of endpoints into bandwidth groups. To add a new bandwidth group use setextradata (this adds a group called Test and sets the maximum to 2 MB/s): VBoxInternal/PDM/AsyncCompletion/File/BwGroups/Test/Max 2097152 VBoxInternal/PDM/AsyncCompletion/File/BwGroups/Test/Start 2097152 VBoxInternal/PDM/AsyncCompletion/File/BwGroups/Test/Step 0 To assign a disk to the group add: VBoxInternal/Devices/ahci/0/LUN#0/AttachedDriver/Config/BwGroup Test

IPRT: RTTcpClientCloseEx - don't be nice to storage servers, they don't always repay the kindness.

Storage/VBoxHDD+iSCSI+DrvVD,Main/Medium: Make the IO and TCPNET interfaces per i mage, fix comments, remove forgotten dead code, handle disk and image interfaces separately in Medium, whitespace cleanup

VBoxHDD: More cleanup - The I/O interface between the generic layer and the backend is private because it includes operations for various async I/O tasks the user of VBoxHDD doesn't know about. Renamed it to make clear that it is internal and make the old async I/O interface the one which can be used for the VFS layer from the outside.

Storage/VBoxHDD: replace custom open flags with regular IPRT file open flags, introduce user-providable filesystem access interface, eliminate dependency on PGM geometry structure, change pvBuffer/cbBuffer parameter ordering to the usual conventions, eliminate the remains of the old I/O code, make more plugin methods optional to reduce redundancy, lots of cleanups Storage/DrvVD+testcases,Main/Medium+Frontends: adapt to VBoxHDD changes, logging fixes Storage/VDI+VMDK+DMG+Raw+VHD+Parallels+VCI: made as similar to each other as possible, added inline VFS wrappers to improve readability, full VFS support, VDI files are now 4K aligned, eliminate the remains of the old I/O code, various more or less severe bugfixes, code sort Storage/iSCSI: support disks bigger than 2T, streamline the code to be more similar to the file-based backends, memory leak fix, error code usage like file-based backends, code sort log+err: added new error codes/log groups and eliminated unused old ones

VD: Little experiment to speed up booting from a high latency storage medium (iSCSI over a WAN network for example) by increasing small reads and saving the result in a temporary buffer. Subsequent reads can then be served from the buffer avoiding to read from the medium.

DrvVD: Implement alternate select interface for Windows XP to avoid the 15ms delay when waiting for the socket to become writable

whitespace

DrvVD: Don't change the events if the socket is not connected

VBoxHDD: Make the non blocking read/write socket APIs available in the VDTCPNET interface

Grrr

VBoxHDD: Add a flag to disable locking of an image file

Main/Medium+Main/Console+Storage: Pass a flag for the medium type "Shareable" to the image format backends so that they can treat the files appropriately.

Main/Console,Storage/DrvVD+fdc: Allow readonly floppy images (suppress runtime error on VM start for readonly media), and take this as the signal to make the floppy medium readonly. Never attempt to write to such images, and signal the appropriate error.

iSCSI: First part for async I/O. Move I/O into a separate thread and handle NOP-in requests properly to prevent disconnects if the guest isn't doing any I/O.

IPRT: - Corrected RTSGBUF member names. - Corrected RTSgBufInit parameter names. - Don't use AssertReturnVoid (and variations) in RTSgBuf* as there the result is that we use uninitialized structures. Better to crash and burn in IPRT then. - Added RTSocketSgWriteL/LV and RTTcpSgWriteL/LV for lazy guys like me. - RTSocketSgWrite doesn't need a do {} while (0).

Storage/iSCSI: switch to sending in a single call, avoiding separate sending of iSCSI PDU header separately on some IP stacks. Adjustments to VBoxHDD, DrvVD and MediumImpl to pass the required VD interface methods.

DrvVD: todo/r=bird: lwip_recv does not set errno, so checking it here is bogus. someone else need to figure this out.

iprt/asm.h,*: Revised the ASMAtomic*Ptr functions and macros. The new saves lots of unsafe (void * volatile *) casts as well as adding some type safety when using GCC (typeof rulez).

Storage/DrvVD: quick workaround for the synchronization problems when live merging a disk which uses async I/O.

VBoxHDD: Move the FinishWrite call from DrvVD to VBoxHDD where we can determine the type of the operation (read/write)

Main: live snapshot merging. initially limited to forward merging (i.e. everything but the first snapshot can be deleted)

Automated rebranding to Oracle copyright/license strings via filemuncher

Interface changes for error reporting and async I/O

oops

VBoxHDD: async I/O updates. Mostly complete except for bug fixes

async I/O: Add async flush method

PDM critsects for drivers. Fixed critsect cleanup in failure path. Started on new transmit locking scheme (required for intnet buffer serialization).

Storage: Convert from PDMDATASEG to RTSGSEG to avoid casting between those two in VBoxHDD and more async I/O updates

VBoxHDD: Async I/O for flat images are back

*: scm cleans up whitespace and adds a new line at the end of ApplianceimplPrivate.h.

DrvVD: Pass request status code to the backends

AsyncCompletion: Return error code for completed requests

Another fix

Compile fix

VBoxHDD: First part of the async I/O support * The async I/O interface is not presented to the backend directly anymore but goes through a new interface in VBoxHDD doing the request management * Implemented everything to make the old coe work again * Async I/O completely disabled for now because it is not working atm (the old code for VMDK was removed)

Storage/VBoxHDD+DrvBlock+DrvVD: implement core code for live snapshot merging.

DrvVD: Implement set size operation for the async I/O interface

DrvVD: Not yet

IPRT: RIP RTCache*, RTMemCache* is way better...

Storage/VBoxHDD+DrvVD: implement framework for providing thread synchronization. Additionally some cleanup to resolve a few minor long-standing todos.

VBoxHDD: introduce new function pointers to the TCP interface for getting the local and peer address of a socket

Devices: warnings.

Don't enable new I/O by default yet

PDM: s/pCfgHandle/pCfg/g - part 2.

PDM: s/szDriverName/szName/g - PDMDRVREG.

AsyncCompletion: Fix incorrect count of outstanding write tasks. Fixes hangs during flush requests. Return VINF_AIO_TASK_PENDING if data needs to be read from the file

PDM,*: Redid the PDM structure versions. Check the instance and helper versions in every device and driver constructor.

pdmifs.h: the final batch of refactored interface ID code.

pdmifs.h: another batch of _IID changes.

PDMIBASE refactoring; use UUID as interface IDs.

PDMDrv,*: multi context drivers, part 2.

PDMDrv,*: multi context drivers, part 1.

*: Use RTMSINTERVAL for timeouts.

AsyncCompletion: Make it possible to enable the new I/O path with a CFGM key. The RTFile* API is still used if the key is not present

DrvVD: Added a LoadDone hook for implementing better handling of TempReadOnly errors (for teleportation).

DrvVD,ConsoleImpl2.cpp: Added a TempReadOnly argument for use on the teleporation target so we can teleport using images without hitting VERR_SHARING_VIOLATION.

DrvVD: Fix stupid bug which masks errors while opening a file

DrvVD.cpp: Fixed file handle leak when taking snapshots.

Storage: Convert the backends to use the interface for I/O rather than using the RTFile* API directly. For a VM the PDMAsyncCompletion is used now as the I/O layer (fixing the I/O load problems). Still disabled by default until it was tested on all supported platforms to not risk corruption of images. To enable compile with VBOX_WITH_NEW_IO_CODE set.

PDMDRVREG change (big changeset).

Storage/VBoxHDD: resurrect the facility to dump information about disk images, and bare minimum fix for creating diff images.

*: s/RT_\(BEGIN|END\)_DECLS/RT_C_DECLS_\1/g

Devices: __BEGIN_DECLS -> RT_BEGIN_DECLS; __END_DECLS -> RT_END_DECLS.

minor compiler warnings

OSE fix

First part of the PDMAsyncCompletion rewrite: * The Host specific parts in VMM are deleted (moved into Runtime) * Implemented a incomplete manager for file operations: - A slightly tested failsafe method to do operations is implemented using the synchronous RTFile* API. The failsafe method will be always used if the host has no support for async I/O or something bad happens during the operation reducing the risk of VM termination * Changed the testcase and drivers to use the new API * Removed the now obsolete async transport driver * Removed the obsolete raw async image driver * Removed the fault injection driver (useless from the beginning)

Storage/DrvVD: don't pass VMSETRTERR_FLAGS_FATAL to ''pfnVMSetRuntimeErrorV()'' as it would cause a deadlock with EMT

Storage/DrvVD: switch to runtime error signalling at the end of driver construction.

DrvVD: fix r44820 for the snapshot case

DrvVD: fail if a VD image was forced to be opened in read-only mode

OSE: export iSCSI initiator, plus minor ifdef cleanup

DrvVD.cpp: doxygen comment in code.

Storage/VBoxHDD: eliminate the obsolete "-new" part of the name.

Main & DrvVD: fix a whole bunch of incorrect CFGM key updates, required to get iSCSI snapshots working.

Storage: Eradicated the last bits using the old VDI only backend, keeping only the testcases for now (no longer built). Completely removed old iSCSI driver. Added intnet option to addiscsidisk and adjusted documentation. Made backend name comparisons case-insensitive. Detect VMDK files not according to VMDK 1.0 and reject with clear error message. Changed format probing logic to not fall through to the "unsupported" case if it's a known format, i.e. has valid header. VBoxManage converthd generic format converter made official. Added format flag to VBoxManage createhd, allows creating VMDK files. VBoxManage convertdd reimplemented based on new framework, supporting any image format. VBoxManage internalcommands sethduuid reimplemented based on new framework, supporting any image format. Cleaned up error codes.

Storage/DrvVD: fix status codes to match the function specs.

VBoxHDD: simplify config handling. Treat everything as a string, and do the conversion in the common code. Makes the job of implementing the VBox API (iSCSI in particular) easier.

DrvVD: fix internal networking IP stuff, not available in OSE.

Device/Storage: fix rc update issue, causing failures for no reason.

Devices/Storage: iSCSI stuff, including bringing IntNet support back to life. Should build, but otherwise untested.

Hex format types (Vhx[sd] -> Rhx[sd]).

Storage: Replace the assertions in the construction code with runtime errors because it is possible that hard disks are not accessible. Also fix a assertion because of a unitiallized semaphore in the IDE controller code

AssertMsgRC

DrvVD: file header.

Better fix for not working VMDK's with SATA, a missing define in the Main Makefile prevented async I/O. More checks in drvvdConstruct.

Failures because rc is not reset. (#3175)

Return error in case of failure...

Storage/DrvVD: warning fix

Storage/VBoxHDD-new: introduced VD interfaces per image and per operation, completely unifying callback handling.

Storage: big cleanup of the VD interfaces, especially hide the linked list better.

Devices: %Vuuid -> %RTuuid (just preferred, not mandatory (yet))

Devices: %Vrc -> %Rrc (just preferred, not mandatory (yet))

Devices: pData -> pThis.

Devices: PDMINS2DATA -> PDMINS_2_DATA.

Devices: VBOX_SUCCESS/FAILURE -> RT_SUCCESS/FAILURE.

Storage: Temporary integration of configuration interface.

Fix invalid pointer to pDrvIns in DrvVD

Fix crashers when destructing DrVD

Merge async I/O for VMDK backend from private branch

The Big Sun Rebranding Header Change