blob: 06ef62aa1c2f6ff3111fd364809605e161f1b706 [file] [log] [blame]
* \file device-flags.h
* Special device flags to deal with bugs in specific devices.
* Copyright (C) 2005-2007 Richard A. Low <>
* Copyright (C) 2005-2012 Linus Walleij <>
* Copyright (C) 2006-2007 Marcus Meissner
* Copyright (C) 2007 Ted Bullock
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* Lesser General Public License for more details.
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
* This file is supposed to be included by both libmtp and libgphoto2.
* These flags are used to indicate if some or other
* device need special treatment. These should be possible
* to concatenate using logical OR so please use one bit per
* feature and lets pray we don't need more than 32 bits...
#define DEVICE_FLAG_NONE 0x00000000
* This means that the PTP_OC_MTP_GetObjPropList is broken
* in the sense that it won't return properly formatted metadata
* for ALL files on the device when you request an object
* property list for object 0xFFFFFFFF with parameter 3 likewise
* set to 0xFFFFFFFF. Compare to
* that it's broken when getting metadata for a SINGLE object.
* A typical way the implementation may be broken is that it
* may not return a proper count of the objects, and sometimes
* (like on the ZENs) objects are simply missing from the list
* if you use this. Sometimes it has been used incorrectly to
* mask bugs in the code (like handling transactions of data
* with size given to -1 (0xFFFFFFFFU), in that case please
* help us remove it now the code is fixed. Sometimes this is
* used because getting all the objects is just too slow and
* the USB transaction will time out if you use this command.
* This means that under Linux, another kernel module may
* be using this device's USB interface, so we need to detach
* it if it is. Typically this is on dual-mode devices that
* will present both an MTP compliant interface and device
* descriptor *and* a USB mass storage interface. If the USB
* mass storage interface is in use, other apps (like our
* userspace libmtp through libusb access path) cannot get in
* and get cosy with it. So we can remove the offending
* application. Typically this means you have to run the program
* as root as well.
#define DEVICE_FLAG_UNLOAD_DRIVER 0x00000002
* This means that the PTP_OC_MTP_GetObjPropList (9805)
* is broken in some way, either it doesn't work at all
* (as for Android devices) or it won't properly return all
* object properties if parameter 3 is set to 0xFFFFFFFFU.
* This means the device doesn't send zero packets to indicate
* end of transfer when the transfer boundary occurs at a
* multiple of 64 bytes (the USB 1.1 endpoint size). Instead,
* exactly one extra byte is sent at the end of the transfer
* if the size is an integer multiple of USB 1.1 endpoint size
* (64 bytes).
* This behaviour is most probably a workaround due to the fact
* that the hardware USB slave controller in the device cannot
* handle zero writes at all, and the usage of the USB 1.1
* endpoint size is due to the fact that the device will "gear
* down" on a USB 1.1 hub, and since 64 bytes is a multiple of
* 512 bytes, it will work with USB 1.1 and USB 2.0 alike.
#define DEVICE_FLAG_NO_ZERO_READS 0x00000008
* This flag means that the device is prone to forgetting the
* OGG container file type, so that libmtp must look at the
* filename extensions in order to determine that a file is
* actually OGG. This is a clear and present firmware bug, and
* while firmware bugs should be fixed in firmware, we like
* OGG so much that we back it by introducing this flag.
* The error has only been seen on iriver devices. Turning this
* flag on won't hurt anything, just that the check against
* filename extension will be done for files of "unknown" type.
* If the player does not even know (reports) that it supports
* ogg even though it does, please use the stronger
* OGG_IS_UNKNOWN flag, which will forcedly support ogg on
* anything with the .ogg filename extension.
* This flag indicates a limitation in the filenames a device
* can accept - they must be 7 bit (all chars <= 127/0x7F).
* It was found first on the Philips Shoqbox, and is a deviation
* from the PTP standard which mandates that any unicode chars
* may be used for filenames. I guess this is caused by a 7bit-only
* filesystem being used intrinsically on the device.
* This flag indicates that the device will lock up if you
* try to get status of endpoints and/or release the interface
* when closing the device. This fixes problems with SanDisk
* Sansa devices especially. It may be a side-effect of a
* Windows behaviour of never releasing interfaces.
* This flag was introduced with the advent of Creative ZEN
* 8GB. The device sometimes return a broken PTP header
* like this: < 1502 0000 0200 01d1 02d1 01d2 >
* the latter 6 bytes (representing "code" and "transaction ID")
* contain junk. This is breaking the PTP/MTP spec but works
* on Windows anyway, probably because the Windows implementation
* does not check that these bytes are valid. To interoperate
* with devices like this, we need this flag to emulate the
* Windows bug. Broken headers has also been found in the
* Aricent MTP stack.
* The Motorola RAZR2 V8 (others?) has broken set object
* proplist causing the metadata setting to fail. (The
* set object prop to set individual properties work on
* this device, but the metadata is plain ignored on
* tracks, though e.g. playlist names can be set.)
* The Samsung YP-T10 think Ogg files shall be sent with
* the "unknown" (PTP_OFC_Undefined) file type, this gives a
* side effect that is a combination of the iRiver Ogg Alzheimer
* problem (have to recognized Ogg files on file extension)
* and a need to report the Ogg support (the device itself does
* not properly claim to support it) and need to set filetype
* to unknown when storing Ogg files, even though they're not
* actually unknown. Later iRivers seem to need this flag since
* they do not report to support OGG even though they actually
* do. Often the device supports OGG in USB mass storage mode,
* then the firmware simply miss to declare metadata support
* for OGG properly.
#define DEVICE_FLAG_OGG_IS_UNKNOWN 0x00000200
* The Creative Zen is quite unstable in libmtp but seems to
* be better with later firmware versions. However, it still
* frequently crashes when setting album art dimensions. This
* flag disables setting the dimensions (which seems to make
* no difference to how the graphic is displayed).
* Some devices, particularly SanDisk Sansas, need to always
* have their "OS Descriptor" probed in order to work correctly.
* This flag provides that extra massage.
* Samsung has implimented its own playlist format as a .spl file
* stored in the normal file system, rather than a proper mtp
* playlist. There are multiple versions of the .spl format
* identified by a line in the file: VERSION X.XX
* Version 1.00 is just a simple playlist.
#define DEVICE_FLAG_PLAYLIST_SPL_V1 0x00001000
* Samsung has implimented its own playlist format as a .spl file
* stored in the normal file system, rather than a proper mtp
* playlist. There are multiple versions of the .spl format
* identified by a line in the file: VERSION X.XX
* Version 2.00 is playlist but allows DNSe sound settings
* to be stored, per playlist.
#define DEVICE_FLAG_PLAYLIST_SPL_V2 0x00002000
* The Sansa E250 is know to have this problem which is actually
* that the device claims that property PTP_OPC_DateModified
* is read/write but will still fail to update it. It can only
* be set properly the first time a file is sent.
* This avoids use of the send object proplist which
* is used when creating new objects (not just updating)
* but only concerns the case where the object proplist
* is sent in to update an existing object. The Toshiba
* Gigabeat MEU202 for example has this problem.
* Devices that cannot support reading out battery
* level.
* Devices that send "ObjectDeleted" events after deletion
* of images. (libgphoto2)
* Cameras that can capture images. (libgphoto2)
#define DEVICE_FLAG_CAPTURE 0x00040000
* Cameras that can capture images. (libgphoto2)
* Nikon broken capture support without proper ObjectAdded events.
* (libgphoto2)
* Broken capture support where cameras do not send CaptureComplete events.
* (libgphoto2)
* Direct PTP match required.
* (libgphoto2)
* This flag is like DEVICE_FLAG_OGG_IS_UNKNOWN but for FLAC
* files instead. Using the unknown filetype for FLAC files.
#define DEVICE_FLAG_FLAC_IS_UNKNOWN 0x01000000
* Device needs unique filenames, no two files can be
* named the same string.
* This flag performs some random magic on the BlackBerry
* device to switch from USB mass storage to MTP mode we think.
* This flag indicates that the device need an extra long
* timeout on some operations.
#define DEVICE_FLAG_LONG_TIMEOUT 0x08000000
* This flag indicates that the device need an explicit
* USB reset after each connection. Some devices don't
* like this, so it's not done by default.
* Early Creative Zen (etc) models actually only support
* command 9805 (Get object property list) and will hang
* if you try to get individual properties of an object.
* It seems that some devices return an bad data when
* using the GetObjectInfo operation. So in these cases
* we prefer to override the PTP-compatible object infos
* with the MTP property list.
* For example Some Samsung Galaxy S devices contain an MTP
* stack that present the ObjectInfo in 64 bit instead of
* 32 bit.
* All these bug flags need to be set on SONY NWZ Walkman
* players, and will be autodetected on unknown devices
* by detecting the vendor extension descriptor ""
* All these bug flags need to be set on Android devices,
* they claim to support MTP operations they actually
* cannot handle, especially 9805 (Get object property list).
* These are auto-assigned to devices reporting
* "" in their device extension descriptor.
* All these bug flags appear on a number of SonyEricsson
* devices including Android devices not using the stock
* Android 4.0+ (Ice Cream Sandwich) MTP stack. It is highly
* supected that these bugs comes from an MTP implementation
* from Aricent, so it is called the Aricent bug flags as a
* shorthand. Especially the header errors that need to be
* ignored is typical for this stack.
* After some guesswork we auto-assign these bug flags to
* devices that present the "", and
* "" but NOT the ""
* descriptor.