2003-06-29 16:33:04 +00:00
|
|
|
/***************************************************************************
|
|
|
|
|
* __________ __ ___.
|
|
|
|
|
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
|
|
|
|
|
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
|
|
|
|
|
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
|
|
|
|
|
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
|
|
|
|
|
* \/ \/ \/ \/ \/
|
|
|
|
|
* $Id$
|
|
|
|
|
*
|
2022-04-10 04:29:01 +00:00
|
|
|
* Copyright (C) 2002 Björn Stenberg
|
2003-06-29 16:33:04 +00:00
|
|
|
*
|
2008-06-28 18:10:04 +00:00
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
|
* modify it under the terms of the GNU General Public License
|
|
|
|
|
* as published by the Free Software Foundation; either version 2
|
|
|
|
|
* of the License, or (at your option) any later version.
|
2003-06-29 16:33:04 +00:00
|
|
|
*
|
|
|
|
|
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
|
|
|
|
|
* KIND, either express or implied.
|
|
|
|
|
*
|
|
|
|
|
****************************************************************************/
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
#define DIRFUNCTIONS_DEFINED
|
|
|
|
|
#define FILEFUNCTIONS_DEFINED
|
2006-11-11 05:11:05 +00:00
|
|
|
#include "plugin.h"
|
2020-08-01 02:45:10 +00:00
|
|
|
#include "open_plugin.h"
|
2006-11-11 05:11:05 +00:00
|
|
|
#include <ctype.h>
|
2003-06-29 16:33:04 +00:00
|
|
|
#include <string.h>
|
2008-04-28 16:18:04 +00:00
|
|
|
#include <stdlib.h>
|
2004-06-10 13:29:52 +00:00
|
|
|
#include "debug.h"
|
2006-11-11 05:47:42 +00:00
|
|
|
#include "i2c.h"
|
2003-06-29 16:33:04 +00:00
|
|
|
#include "lang.h"
|
2007-07-29 03:43:37 +00:00
|
|
|
#include "led.h"
|
2003-07-09 22:36:23 +00:00
|
|
|
#include "keyboard.h"
|
2004-01-30 22:30:40 +00:00
|
|
|
#include "backlight.h"
|
2006-03-26 22:33:25 +00:00
|
|
|
#include "sound_menu.h"
|
2004-06-10 13:29:52 +00:00
|
|
|
#include "mp3data.h"
|
2004-06-15 22:54:37 +00:00
|
|
|
#include "powermgmt.h"
|
2005-11-16 15:12:15 +00:00
|
|
|
#include "splash.h"
|
2006-11-11 08:00:28 +00:00
|
|
|
#include "logf.h"
|
2007-07-11 05:41:23 +00:00
|
|
|
#include "option_select.h"
|
2007-10-03 16:17:36 +00:00
|
|
|
#include "talk.h"
|
2008-09-22 19:59:00 +00:00
|
|
|
#include "version.h"
|
2008-11-01 16:14:28 +00:00
|
|
|
#include "storage.h"
|
2009-10-29 16:52:12 +00:00
|
|
|
#include "pcmbuf.h"
|
2009-12-24 17:07:20 +00:00
|
|
|
#include "errno.h"
|
2010-04-07 17:22:16 +00:00
|
|
|
#include "diacritic.h"
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
#include "pathfuncs.h"
|
2010-08-27 00:29:50 +00:00
|
|
|
#include "load_code.h"
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
#include "file.h"
|
2022-02-24 02:26:37 +00:00
|
|
|
#include "core_keymap.h"
|
2003-07-09 22:36:23 +00:00
|
|
|
|
2007-02-18 05:32:06 +00:00
|
|
|
#if CONFIG_CHARGING
|
2006-02-07 14:46:03 +00:00
|
|
|
#include "power.h"
|
|
|
|
|
#endif
|
|
|
|
|
|
2006-11-19 14:11:42 +00:00
|
|
|
#include "scrollbar.h"
|
2007-04-10 14:18:30 +00:00
|
|
|
#include "peakmeter.h"
|
2005-04-25 07:42:10 +00:00
|
|
|
#include "bmp.h"
|
2006-07-19 19:47:34 +00:00
|
|
|
#include "bidi.h"
|
2003-06-29 16:33:04 +00:00
|
|
|
|
2009-09-30 10:06:04 +00:00
|
|
|
#ifdef USB_ENABLE_HID
|
2009-07-18 12:58:29 +00:00
|
|
|
#include "usbstack/usb_hid.h"
|
|
|
|
|
#endif
|
|
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
#define WRAPPER(_x_) _x_ ## _wrapper
|
|
|
|
|
|
|
|
|
|
#if (CONFIG_PLATFORM & PLATFORM_HOSTED)
|
|
|
|
|
static unsigned char pluginbuf[PLUGIN_BUFFER_SIZE];
|
|
|
|
|
void sim_lcd_ex_init(unsigned long (*getpixel)(int, int));
|
|
|
|
|
void sim_lcd_ex_update_rect(int x, int y, int width, int height);
|
2009-01-07 09:53:46 +00:00
|
|
|
#else
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
extern unsigned char pluginbuf[];
|
|
|
|
|
#include "bitswap.h"
|
2009-01-07 09:53:46 +00:00
|
|
|
#endif
|
|
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
/* for actual plugins only, not for codecs */
|
|
|
|
|
static int plugin_size = 0;
|
2023-03-21 02:15:33 +00:00
|
|
|
static int (*pfn_tsr_exit)(bool reenter) = NULL; /* TSR exit callback */
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static char current_plugin[MAX_PATH];
|
|
|
|
|
/* NULL if no plugin is loaded, otherwise the handle that lc_open() returned */
|
|
|
|
|
static void *current_plugin_handle;
|
|
|
|
|
|
|
|
|
|
char *plugin_get_current_filename(void);
|
|
|
|
|
|
|
|
|
|
static void* plugin_get_audio_buffer(size_t *buffer_size);
|
|
|
|
|
static void plugin_release_audio_buffer(void);
|
2023-03-21 02:15:33 +00:00
|
|
|
static void plugin_tsr(int (*exit_callback)(bool));
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
|
|
|
|
|
#ifdef HAVE_PLUGIN_CHECK_OPEN_CLOSE
|
|
|
|
|
/* File handle leak prophylaxis */
|
|
|
|
|
#include "bitarray.h"
|
|
|
|
|
#include "file_internal.h" /* for MAX_OPEN_FILES */
|
|
|
|
|
|
|
|
|
|
#define PCOC_WRAPPER(_x_) WRAPPER(_x_)
|
|
|
|
|
|
|
|
|
|
BITARRAY_TYPE_DECLARE(plugin_check_open_close_bitmap_t, open_files_bitmap,
|
|
|
|
|
MAX_OPEN_FILES)
|
|
|
|
|
|
|
|
|
|
static plugin_check_open_close_bitmap_t open_files_bitmap;
|
|
|
|
|
|
|
|
|
|
static void plugin_check_open_close__enter(void)
|
2011-01-07 18:45:35 +00:00
|
|
|
{
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
if (!current_plugin_handle)
|
|
|
|
|
open_files_bitmap_clear(&open_files_bitmap);
|
2011-01-07 18:45:35 +00:00
|
|
|
}
|
|
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static void plugin_check_open_close__open(int fildes)
|
2011-01-07 18:45:35 +00:00
|
|
|
{
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
if (fildes >= 0)
|
|
|
|
|
open_files_bitmap_set_bit(&open_files_bitmap, fildes);
|
2011-01-07 18:45:35 +00:00
|
|
|
}
|
|
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static void plugin_check_open_close__close(int fildes)
|
2011-01-07 18:45:35 +00:00
|
|
|
{
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
if (fildes < 0)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
if (!open_files_bitmap_test_bit(&open_files_bitmap, fildes))
|
|
|
|
|
{
|
|
|
|
|
logf("double close from plugin");
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
open_files_bitmap_clear_bit(&open_files_bitmap, fildes);
|
2011-01-07 18:45:35 +00:00
|
|
|
}
|
|
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static int WRAPPER(open)(const char *path, int oflag, ...)
|
2011-01-07 18:45:35 +00:00
|
|
|
{
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
int fildes = FS_PREFIX(open)(path, oflag __OPEN_MODE_ARG);
|
|
|
|
|
plugin_check_open_close__open(fildes);
|
|
|
|
|
return fildes;
|
2011-01-07 18:45:35 +00:00
|
|
|
}
|
|
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static int WRAPPER(creat)(const char *path, mode_t mode)
|
2011-01-07 18:45:35 +00:00
|
|
|
{
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
int fildes = FS_PREFIX(creat)(path __CREAT_MODE_ARG);
|
|
|
|
|
plugin_check_open_close__open(fildes);
|
|
|
|
|
return fildes;
|
|
|
|
|
(void)mode;
|
2011-01-07 18:45:35 +00:00
|
|
|
}
|
|
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static int WRAPPER(close)(int fildes)
|
|
|
|
|
{
|
|
|
|
|
int rc = FS_PREFIX(close)(fildes);
|
|
|
|
|
if (rc >= 0)
|
|
|
|
|
plugin_check_open_close__close(fildes);
|
2010-02-11 19:15:37 +00:00
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
return rc;
|
|
|
|
|
}
|
2010-02-11 19:15:37 +00:00
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static void plugin_check_open_close__exit(void)
|
|
|
|
|
{
|
|
|
|
|
if (current_plugin_handle)
|
|
|
|
|
return;
|
2003-07-25 23:04:59 +00:00
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
if (open_files_bitmap_is_clear(&open_files_bitmap))
|
|
|
|
|
return;
|
2003-07-25 23:04:59 +00:00
|
|
|
|
2016-03-30 20:48:17 +00:00
|
|
|
logf("Plugin '%s' leaks file handles", current_plugin);
|
2007-08-06 13:42:52 +00:00
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
static const char *lines[] =
|
|
|
|
|
{ ID2P(LANG_PLUGIN_ERROR), "#leak-file-handles" };
|
|
|
|
|
static const struct text_message message = { lines, 2 };
|
|
|
|
|
button_clear_queue(); /* Empty the keyboard buffer */
|
|
|
|
|
gui_syncyesno_run(&message, NULL, NULL);
|
2010-02-11 19:15:37 +00:00
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
FOR_EACH_BITARRAY_SET_BIT(&open_files_bitmap, fildes)
|
|
|
|
|
WRAPPER(close)(fildes);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#else /* !HAVE_PLUGIN_CHECK_OPEN_CLOSE */
|
|
|
|
|
|
|
|
|
|
#define PCOC_WRAPPER(_x_) FS_PREFIX(_x_)
|
|
|
|
|
#define plugin_check_open_close__enter()
|
|
|
|
|
#define plugin_check_open_close__exit()
|
|
|
|
|
|
|
|
|
|
#endif /* HAVE_PLUGIN_CHECK_OPEN_CLOSE */
|
2014-04-09 07:00:36 +00:00
|
|
|
|
2004-07-20 21:37:36 +00:00
|
|
|
static const struct plugin_api rockbox_api = {
|
2019-02-04 01:12:50 +00:00
|
|
|
rbversion,
|
|
|
|
|
&global_settings,
|
|
|
|
|
&global_status,
|
|
|
|
|
language_strings,
|
2021-08-26 13:37:31 +00:00
|
|
|
&core_bitmaps[0],
|
2003-06-29 16:33:04 +00:00
|
|
|
/* lcd */
|
2019-07-20 09:46:14 +00:00
|
|
|
splash,
|
|
|
|
|
splashf,
|
2023-03-21 02:15:33 +00:00
|
|
|
splash_progress,
|
|
|
|
|
splash_progress_set_delay,
|
2009-03-30 16:06:02 +00:00
|
|
|
#ifdef HAVE_LCD_CONTRAST
|
2005-06-24 22:33:21 +00:00
|
|
|
lcd_set_contrast,
|
2009-03-30 16:06:02 +00:00
|
|
|
#endif
|
2007-03-31 09:58:49 +00:00
|
|
|
lcd_update,
|
2003-06-29 16:33:04 +00:00
|
|
|
lcd_clear_display,
|
2007-03-26 07:52:13 +00:00
|
|
|
lcd_getstringsize,
|
|
|
|
|
lcd_putsxy,
|
2010-08-30 20:47:53 +00:00
|
|
|
lcd_putsxyf,
|
2003-06-29 16:33:04 +00:00
|
|
|
lcd_puts,
|
2010-08-30 20:47:53 +00:00
|
|
|
lcd_putsf,
|
2003-06-29 16:33:04 +00:00
|
|
|
lcd_puts_scroll,
|
2013-04-03 14:33:23 +00:00
|
|
|
lcd_scroll_stop,
|
2012-12-09 20:04:17 +00:00
|
|
|
lcd_set_viewport,
|
|
|
|
|
lcd_bmp_part,
|
2008-03-24 00:45:03 +00:00
|
|
|
lcd_update_rect,
|
2005-06-24 22:33:21 +00:00
|
|
|
lcd_set_drawmode,
|
|
|
|
|
lcd_get_drawmode,
|
2011-09-24 13:19:34 +00:00
|
|
|
screen_helper_setfont,
|
2005-06-24 22:33:21 +00:00
|
|
|
lcd_drawpixel,
|
|
|
|
|
lcd_drawline,
|
2005-06-28 23:15:47 +00:00
|
|
|
lcd_hline,
|
|
|
|
|
lcd_vline,
|
2005-06-24 22:33:21 +00:00
|
|
|
lcd_drawrect,
|
|
|
|
|
lcd_fillrect,
|
2005-07-06 22:58:02 +00:00
|
|
|
lcd_mono_bitmap_part,
|
|
|
|
|
lcd_mono_bitmap,
|
2005-07-26 20:01:11 +00:00
|
|
|
#if LCD_DEPTH > 1
|
|
|
|
|
lcd_set_foreground,
|
|
|
|
|
lcd_get_foreground,
|
|
|
|
|
lcd_set_background,
|
|
|
|
|
lcd_get_background,
|
|
|
|
|
lcd_bitmap_part,
|
|
|
|
|
lcd_bitmap,
|
2007-06-22 09:34:57 +00:00
|
|
|
lcd_get_backdrop,
|
2006-11-19 14:11:42 +00:00
|
|
|
lcd_set_backdrop,
|
2006-02-07 20:38:55 +00:00
|
|
|
#endif
|
2014-06-18 05:15:00 +00:00
|
|
|
#if LCD_DEPTH >= 16
|
2006-02-07 20:38:55 +00:00
|
|
|
lcd_bitmap_transparent_part,
|
|
|
|
|
lcd_bitmap_transparent,
|
2022-10-13 15:03:53 +00:00
|
|
|
#if MEMORYSIZE > 2
|
|
|
|
|
lcd_blit_yuv,
|
|
|
|
|
#if defined(TOSHIBA_GIGABEAT_F) || defined(SANSA_E200) || defined(SANSA_C200) \
|
|
|
|
|
|| defined(IRIVER_H10) || defined(COWON_D2) || defined(PHILIPS_HDD1630) \
|
|
|
|
|
|| defined(SANSA_FUZE) || defined(SANSA_E200V2) || defined(SANSA_FUZEV2) \
|
|
|
|
|
|| defined(TOSHIBA_GIGABEAT_S) || defined(PHILIPS_SA9200)
|
|
|
|
|
lcd_yuv_set_options,
|
|
|
|
|
#endif
|
|
|
|
|
#endif /* MEMORYSIZE > 2 */
|
2010-06-21 16:53:00 +00:00
|
|
|
#elif (LCD_DEPTH < 4) && (CONFIG_PLATFORM & PLATFORM_NATIVE)
|
2008-03-24 00:35:53 +00:00
|
|
|
lcd_blit_mono,
|
|
|
|
|
lcd_blit_grey_phase,
|
|
|
|
|
#endif /* LCD_DEPTH */
|
2009-05-06 04:53:56 +00:00
|
|
|
#if defined(HAVE_LCD_MODES) && (HAVE_LCD_MODES & LCD_MODE_PAL256)
|
|
|
|
|
lcd_blit_pal256,
|
|
|
|
|
lcd_pal256_update_pal,
|
|
|
|
|
#endif
|
2008-05-15 06:58:36 +00:00
|
|
|
#ifdef HAVE_LCD_INVERT
|
|
|
|
|
lcd_set_invert_display,
|
|
|
|
|
#endif /* HAVE_LCD_INVERT */
|
2009-05-06 04:53:56 +00:00
|
|
|
#if defined(HAVE_LCD_MODES)
|
|
|
|
|
lcd_set_mode,
|
|
|
|
|
#endif
|
2009-03-17 02:43:47 +00:00
|
|
|
#if defined(HAVE_LCD_ENABLE) || defined(HAVE_LCD_SLEEP)
|
2008-06-23 13:20:35 +00:00
|
|
|
&button_queue,
|
|
|
|
|
#endif
|
2006-09-16 16:18:11 +00:00
|
|
|
bidi_l2v,
|
2010-05-24 16:42:32 +00:00
|
|
|
is_diacritic,
|
2006-09-16 16:18:11 +00:00
|
|
|
font_get_bits,
|
|
|
|
|
font_load,
|
2011-09-24 13:19:34 +00:00
|
|
|
font_unload,
|
2005-06-24 22:33:21 +00:00
|
|
|
font_get,
|
2005-07-26 20:01:11 +00:00
|
|
|
font_getstringsize,
|
2006-01-15 18:20:18 +00:00
|
|
|
font_get_width,
|
2007-08-06 13:42:52 +00:00
|
|
|
screen_clear_area,
|
2008-03-24 00:35:53 +00:00
|
|
|
gui_scrollbar_draw,
|
2019-07-20 20:06:30 +00:00
|
|
|
get_codepage_name,
|
2008-03-24 00:35:53 +00:00
|
|
|
|
2005-04-15 12:55:31 +00:00
|
|
|
#ifdef HAVE_REMOTE_LCD
|
2005-06-18 01:25:47 +00:00
|
|
|
/* remote lcd */
|
2005-06-29 01:39:50 +00:00
|
|
|
lcd_remote_set_contrast,
|
2005-04-25 07:42:10 +00:00
|
|
|
lcd_remote_clear_display,
|
2005-04-25 23:47:51 +00:00
|
|
|
lcd_remote_puts,
|
|
|
|
|
lcd_remote_puts_scroll,
|
2013-04-03 14:33:23 +00:00
|
|
|
lcd_remote_scroll_stop,
|
2005-06-29 01:39:50 +00:00
|
|
|
lcd_remote_set_drawmode,
|
|
|
|
|
lcd_remote_get_drawmode,
|
|
|
|
|
lcd_remote_setfont,
|
|
|
|
|
lcd_remote_getstringsize,
|
|
|
|
|
lcd_remote_drawpixel,
|
|
|
|
|
lcd_remote_drawline,
|
|
|
|
|
lcd_remote_hline,
|
|
|
|
|
lcd_remote_vline,
|
|
|
|
|
lcd_remote_drawrect,
|
|
|
|
|
lcd_remote_fillrect,
|
2005-07-07 00:05:29 +00:00
|
|
|
lcd_remote_mono_bitmap_part,
|
|
|
|
|
lcd_remote_mono_bitmap,
|
2005-04-25 23:47:51 +00:00
|
|
|
lcd_remote_putsxy,
|
|
|
|
|
lcd_remote_update,
|
|
|
|
|
lcd_remote_update_rect,
|
2019-07-20 09:46:14 +00:00
|
|
|
#if (LCD_REMOTE_DEPTH > 1)
|
2006-09-16 16:18:11 +00:00
|
|
|
lcd_remote_set_foreground,
|
|
|
|
|
lcd_remote_get_foreground,
|
|
|
|
|
lcd_remote_set_background,
|
|
|
|
|
lcd_remote_get_background,
|
|
|
|
|
lcd_remote_bitmap_part,
|
|
|
|
|
lcd_remote_bitmap,
|
|
|
|
|
#endif
|
2019-07-20 09:46:14 +00:00
|
|
|
#endif /* HAVE_REMOTE_LCD */
|
|
|
|
|
#if NB_SCREENS == 2
|
|
|
|
|
{&screens[SCREEN_MAIN], &screens[SCREEN_REMOTE]},
|
|
|
|
|
#else
|
|
|
|
|
{&screens[SCREEN_MAIN]},
|
|
|
|
|
#endif
|
|
|
|
|
|
2008-03-26 03:35:24 +00:00
|
|
|
viewport_set_defaults,
|
2010-02-14 06:26:16 +00:00
|
|
|
viewportmanager_theme_enable,
|
|
|
|
|
viewportmanager_theme_undo,
|
2012-12-09 20:04:17 +00:00
|
|
|
viewport_set_fullscreen,
|
2020-10-07 06:01:35 +00:00
|
|
|
viewport_set_buffer,
|
2019-02-04 01:12:50 +00:00
|
|
|
|
2020-07-15 23:40:55 +00:00
|
|
|
#ifdef HAVE_BACKLIGHT
|
2019-07-20 09:46:14 +00:00
|
|
|
/* lcd backlight */
|
2020-07-15 23:40:55 +00:00
|
|
|
/* For OLED targets like the Sansa Clip, the backlight_* functions control
|
2019-07-20 09:46:14 +00:00
|
|
|
* the display enable, which has essentially the same effect. */
|
|
|
|
|
is_backlight_on,
|
|
|
|
|
backlight_on,
|
|
|
|
|
backlight_off,
|
|
|
|
|
backlight_set_timeout,
|
|
|
|
|
#ifdef HAVE_BACKLIGHT_BRIGHTNESS
|
|
|
|
|
backlight_set_brightness,
|
|
|
|
|
#endif /* HAVE_BACKLIGHT_BRIGHTNESS */
|
|
|
|
|
|
|
|
|
|
#if CONFIG_CHARGING
|
|
|
|
|
backlight_set_timeout_plugged,
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
#ifdef HAVE_REMOTE_LCD
|
|
|
|
|
remote_backlight_on,
|
|
|
|
|
remote_backlight_off,
|
|
|
|
|
remote_backlight_set_timeout,
|
|
|
|
|
#if CONFIG_CHARGING
|
|
|
|
|
remote_backlight_set_timeout_plugged,
|
|
|
|
|
#endif
|
|
|
|
|
#endif /* HAVE_REMOTE_LCD */
|
2020-07-15 23:40:55 +00:00
|
|
|
#endif /* HAVE_BACKLIGHT */
|
2019-07-20 09:46:14 +00:00
|
|
|
|
2006-09-16 16:18:11 +00:00
|
|
|
/* list */
|
|
|
|
|
gui_synclist_init,
|
|
|
|
|
gui_synclist_set_nb_items,
|
2019-02-04 01:12:50 +00:00
|
|
|
gui_synclist_set_voice_callback,
|
2006-09-16 16:18:11 +00:00
|
|
|
gui_synclist_set_icon_callback,
|
|
|
|
|
gui_synclist_get_nb_items,
|
|
|
|
|
gui_synclist_get_sel_pos,
|
|
|
|
|
gui_synclist_draw,
|
2019-02-04 01:12:50 +00:00
|
|
|
gui_synclist_speak_item,
|
2006-09-16 16:18:11 +00:00
|
|
|
gui_synclist_select_item,
|
|
|
|
|
gui_synclist_add_item,
|
|
|
|
|
gui_synclist_del_item,
|
|
|
|
|
gui_synclist_do_button,
|
2006-11-30 22:29:48 +00:00
|
|
|
gui_synclist_set_title,
|
2008-08-14 22:35:00 +00:00
|
|
|
gui_syncyesno_run,
|
2008-06-23 13:20:35 +00:00
|
|
|
simplelist_info_init,
|
|
|
|
|
simplelist_show_list,
|
2021-11-11 05:05:29 +00:00
|
|
|
yesno_pop,
|
2006-10-14 01:32:58 +00:00
|
|
|
|
2019-07-20 09:46:14 +00:00
|
|
|
/* action handling */
|
|
|
|
|
get_custom_action,
|
|
|
|
|
get_action,
|
|
|
|
|
#ifdef HAVE_TOUCHSCREEN
|
|
|
|
|
action_get_touchscreen_press,
|
2021-08-20 02:20:56 +00:00
|
|
|
action_get_touchscreen_press_in_vp,
|
2019-07-20 09:46:14 +00:00
|
|
|
#endif
|
|
|
|
|
action_userabort,
|
2022-09-19 10:10:54 +00:00
|
|
|
core_set_keyremap,
|
2019-07-20 09:46:14 +00:00
|
|
|
|
2003-06-29 16:33:04 +00:00
|
|
|
/* button */
|
|
|
|
|
button_get,
|
|
|
|
|
button_get_w_tmo,
|
2004-07-27 14:10:48 +00:00
|
|
|
button_status,
|
2008-05-15 06:58:36 +00:00
|
|
|
#ifdef HAVE_BUTTON_DATA
|
|
|
|
|
button_get_data,
|
2009-10-20 21:54:59 +00:00
|
|
|
button_status_wdata,
|
2008-05-15 06:58:36 +00:00
|
|
|
#endif
|
2004-07-27 14:10:48 +00:00
|
|
|
button_clear_queue,
|
2007-10-16 01:25:17 +00:00
|
|
|
button_queue_count,
|
2006-05-01 05:45:18 +00:00
|
|
|
#ifdef HAS_BUTTON_HOLD
|
2005-06-18 01:25:47 +00:00
|
|
|
button_hold,
|
|
|
|
|
#endif
|
2020-06-27 00:53:15 +00:00
|
|
|
#ifdef HAVE_SW_POWEROFF
|
|
|
|
|
button_set_sw_poweroff_state,
|
|
|
|
|
button_get_sw_poweroff_state,
|
|
|
|
|
#endif
|
2008-08-23 09:46:38 +00:00
|
|
|
#ifdef HAVE_TOUCHSCREEN
|
|
|
|
|
touchscreen_set_mode,
|
2012-01-26 22:37:27 +00:00
|
|
|
touchscreen_get_mode,
|
2008-05-15 06:58:36 +00:00
|
|
|
#endif
|
2019-02-04 01:12:50 +00:00
|
|
|
|
2008-05-15 06:58:36 +00:00
|
|
|
#ifdef HAVE_BUTTON_LIGHT
|
|
|
|
|
buttonlight_set_timeout,
|
|
|
|
|
buttonlight_off,
|
|
|
|
|
buttonlight_on,
|
|
|
|
|
#ifdef HAVE_BUTTONLIGHT_BRIGHTNESS
|
|
|
|
|
buttonlight_set_brightness,
|
|
|
|
|
#endif /* HAVE_BUTTONLIGHT_BRIGHTNESS */
|
|
|
|
|
#endif /* HAVE_BUTTON_LIGHT */
|
2003-06-29 16:33:04 +00:00
|
|
|
|
|
|
|
|
/* file */
|
2010-02-14 06:26:16 +00:00
|
|
|
open_utf8,
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
PCOC_WRAPPER(open),
|
|
|
|
|
PCOC_WRAPPER(creat),
|
|
|
|
|
PCOC_WRAPPER(close),
|
|
|
|
|
FS_PREFIX(read),
|
|
|
|
|
FS_PREFIX(lseek),
|
|
|
|
|
FS_PREFIX(write),
|
|
|
|
|
FS_PREFIX(remove),
|
|
|
|
|
FS_PREFIX(rename),
|
|
|
|
|
FS_PREFIX(ftruncate),
|
|
|
|
|
FS_PREFIX(filesize),
|
2005-02-22 12:19:12 +00:00
|
|
|
fdprintf,
|
2003-06-29 16:33:04 +00:00
|
|
|
read_line,
|
2004-07-27 14:10:48 +00:00
|
|
|
settings_parseline,
|
2008-11-01 16:14:28 +00:00
|
|
|
storage_sleep,
|
2020-07-17 03:23:25 +00:00
|
|
|
STORAGE_FUNCTION(spin),
|
|
|
|
|
STORAGE_FUNCTION(spindown),
|
2008-11-01 16:14:28 +00:00
|
|
|
#if USING_STORAGE_CALLBACK
|
|
|
|
|
register_storage_idle_func,
|
|
|
|
|
unregister_storage_idle_func,
|
|
|
|
|
#endif /* USING_STORAGE_CALLBACK */
|
2006-09-16 16:18:11 +00:00
|
|
|
reload_directory,
|
2007-06-22 09:34:57 +00:00
|
|
|
create_numbered_filename,
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
FS_PREFIX(file_exists),
|
2009-02-11 16:37:12 +00:00
|
|
|
strip_extension,
|
2010-02-14 06:26:16 +00:00
|
|
|
crc_32,
|
2021-08-20 02:20:56 +00:00
|
|
|
crc_32r,
|
2011-01-20 17:11:23 +00:00
|
|
|
filetype_get_attr,
|
2003-06-29 16:33:04 +00:00
|
|
|
|
|
|
|
|
/* dir */
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
FS_PREFIX(opendir),
|
|
|
|
|
FS_PREFIX(closedir),
|
|
|
|
|
FS_PREFIX(readdir),
|
|
|
|
|
FS_PREFIX(mkdir),
|
|
|
|
|
FS_PREFIX(rmdir),
|
|
|
|
|
FS_PREFIX(dir_exists),
|
2010-10-16 12:30:06 +00:00
|
|
|
dir_get_info,
|
2003-06-29 16:33:04 +00:00
|
|
|
|
2011-01-20 17:11:23 +00:00
|
|
|
/* browsing */
|
|
|
|
|
rockbox_browse,
|
2019-07-20 09:46:14 +00:00
|
|
|
tree_get_context,
|
|
|
|
|
tree_get_entries,
|
|
|
|
|
tree_get_entry_at,
|
|
|
|
|
set_current_file,
|
|
|
|
|
set_dirfilter,
|
2022-09-19 10:10:54 +00:00
|
|
|
onplay_show_playlist_menu,
|
2023-03-21 02:15:33 +00:00
|
|
|
onplay_show_playlist_cat_menu,
|
2022-09-19 10:10:54 +00:00
|
|
|
browse_id3,
|
2011-01-20 17:11:23 +00:00
|
|
|
|
2019-02-04 01:12:50 +00:00
|
|
|
/* talking */
|
|
|
|
|
talk_id,
|
|
|
|
|
talk_file,
|
|
|
|
|
talk_file_or_spell,
|
|
|
|
|
talk_dir_or_spell,
|
|
|
|
|
talk_number,
|
2021-08-18 05:23:26 +00:00
|
|
|
talk_value_decimal,
|
2019-02-04 01:12:50 +00:00
|
|
|
talk_spell,
|
|
|
|
|
talk_time,
|
|
|
|
|
talk_date,
|
|
|
|
|
talk_disable,
|
|
|
|
|
talk_shutup,
|
|
|
|
|
talk_force_shutup,
|
|
|
|
|
talk_force_enqueue_next,
|
|
|
|
|
|
2004-10-23 09:59:53 +00:00
|
|
|
/* kernel/ system */
|
2010-08-03 19:00:29 +00:00
|
|
|
#if defined(CPU_ARM) && CONFIG_PLATFORM & PLATFORM_NATIVE
|
2009-02-11 16:37:12 +00:00
|
|
|
__div0,
|
|
|
|
|
#endif
|
2010-09-01 21:29:34 +00:00
|
|
|
sleep,
|
2004-07-27 14:10:48 +00:00
|
|
|
yield,
|
2003-06-29 16:33:04 +00:00
|
|
|
¤t_tick,
|
2004-07-27 14:10:48 +00:00
|
|
|
default_event_handler,
|
2004-10-23 09:59:53 +00:00
|
|
|
default_event_handler_ex,
|
2004-07-27 14:10:48 +00:00
|
|
|
create_thread,
|
2011-03-16 05:38:37 +00:00
|
|
|
thread_self,
|
2008-03-25 02:34:12 +00:00
|
|
|
thread_exit,
|
2008-03-24 00:35:53 +00:00
|
|
|
thread_wait,
|
2009-01-05 10:31:19 +00:00
|
|
|
thread_thaw,
|
|
|
|
|
#ifdef HAVE_PRIORITY_SCHEDULING
|
|
|
|
|
thread_set_priority,
|
|
|
|
|
#endif
|
2008-03-24 00:35:53 +00:00
|
|
|
mutex_init,
|
|
|
|
|
mutex_lock,
|
|
|
|
|
mutex_unlock,
|
2019-07-20 09:46:14 +00:00
|
|
|
#ifdef HAVE_SEMAPHORE_OBJECTS
|
|
|
|
|
semaphore_init,
|
|
|
|
|
semaphore_wait,
|
|
|
|
|
semaphore_release,
|
|
|
|
|
#endif
|
2004-07-27 19:16:58 +00:00
|
|
|
reset_poweroff_timer,
|
2019-07-20 09:46:14 +00:00
|
|
|
set_sleeptimer_duration, /*stub*/
|
|
|
|
|
get_sleep_timer, /*stub*/
|
2010-06-21 16:53:00 +00:00
|
|
|
#if (CONFIG_PLATFORM & PLATFORM_NATIVE)
|
2004-10-23 09:59:53 +00:00
|
|
|
system_memory_guard,
|
2005-06-18 01:25:47 +00:00
|
|
|
&cpu_frequency,
|
2007-01-22 10:41:25 +00:00
|
|
|
|
2005-06-18 01:25:47 +00:00
|
|
|
#ifdef HAVE_ADJUSTABLE_CPU_FREQ
|
2007-01-22 10:41:25 +00:00
|
|
|
#ifdef CPU_BOOST_LOGGING
|
|
|
|
|
cpu_boost_,
|
|
|
|
|
#else
|
2005-06-18 01:25:47 +00:00
|
|
|
cpu_boost,
|
2007-01-22 10:41:25 +00:00
|
|
|
#endif
|
2008-03-24 00:35:53 +00:00
|
|
|
#endif /* HAVE_ADJUSTABLE_CPU_FREQ */
|
2010-06-21 16:53:00 +00:00
|
|
|
#endif /* PLATFORM_NATIVE */
|
2008-03-24 00:35:53 +00:00
|
|
|
#ifdef HAVE_SCHEDULER_BOOSTCTRL
|
|
|
|
|
trigger_cpu_boost,
|
|
|
|
|
cancel_cpu_boost,
|
2007-07-29 03:43:37 +00:00
|
|
|
#endif
|
2010-10-16 12:30:06 +00:00
|
|
|
|
2011-12-17 07:27:24 +00:00
|
|
|
commit_dcache,
|
|
|
|
|
commit_discard_dcache,
|
2012-01-27 08:31:34 +00:00
|
|
|
commit_discard_idcache,
|
2010-10-16 12:30:06 +00:00
|
|
|
|
|
|
|
|
lc_open,
|
|
|
|
|
lc_open_from_mem,
|
|
|
|
|
lc_get_header,
|
|
|
|
|
lc_close,
|
|
|
|
|
|
2005-07-26 20:01:11 +00:00
|
|
|
timer_register,
|
|
|
|
|
timer_unregister,
|
|
|
|
|
timer_set_period,
|
2007-07-29 03:43:37 +00:00
|
|
|
|
2006-02-07 20:38:55 +00:00
|
|
|
queue_init,
|
|
|
|
|
queue_delete,
|
|
|
|
|
queue_post,
|
|
|
|
|
queue_wait_w_tmo,
|
2008-03-24 00:35:53 +00:00
|
|
|
queue_enable_queue_send,
|
|
|
|
|
queue_empty,
|
|
|
|
|
queue_wait,
|
|
|
|
|
queue_send,
|
|
|
|
|
queue_reply,
|
2022-09-19 10:10:54 +00:00
|
|
|
queue_remove_from_head,
|
2019-07-20 09:46:14 +00:00
|
|
|
|
2006-02-07 20:38:55 +00:00
|
|
|
#ifdef RB_PROFILE
|
|
|
|
|
profile_thread,
|
|
|
|
|
profstop,
|
2008-07-28 22:43:41 +00:00
|
|
|
__cyg_profile_func_enter,
|
|
|
|
|
__cyg_profile_func_exit,
|
2006-02-07 20:38:55 +00:00
|
|
|
#endif
|
2009-10-20 21:54:59 +00:00
|
|
|
add_event,
|
|
|
|
|
remove_event,
|
|
|
|
|
send_event,
|
2003-06-29 16:33:04 +00:00
|
|
|
|
2010-06-21 16:53:00 +00:00
|
|
|
#if (CONFIG_PLATFORM & PLATFORM_HOSTED)
|
2006-03-11 09:28:49 +00:00
|
|
|
/* special simulator hooks */
|
2020-07-17 14:31:31 +00:00
|
|
|
#if LCD_DEPTH < 8
|
2006-03-11 09:28:49 +00:00
|
|
|
sim_lcd_ex_init,
|
|
|
|
|
sim_lcd_ex_update_rect,
|
|
|
|
|
#endif
|
|
|
|
|
#endif
|
|
|
|
|
|
2003-06-29 16:33:04 +00:00
|
|
|
/* strings and memory */
|
|
|
|
|
snprintf,
|
2006-09-16 16:18:11 +00:00
|
|
|
vsnprintf,
|
2003-06-29 16:33:04 +00:00
|
|
|
strcpy,
|
2009-07-14 13:57:45 +00:00
|
|
|
strlcpy,
|
2003-06-29 16:33:04 +00:00
|
|
|
strlen,
|
2004-07-13 14:14:30 +00:00
|
|
|
strrchr,
|
2004-07-27 14:10:48 +00:00
|
|
|
strcmp,
|
2005-07-26 20:01:11 +00:00
|
|
|
strncmp,
|
2004-07-27 14:10:48 +00:00
|
|
|
strcasecmp,
|
2004-10-23 09:59:53 +00:00
|
|
|
strncasecmp,
|
2003-06-29 16:33:04 +00:00
|
|
|
memset,
|
|
|
|
|
memcpy,
|
2006-02-07 20:38:55 +00:00
|
|
|
memmove,
|
2010-06-21 16:53:00 +00:00
|
|
|
#if (CONFIG_PLATFORM & PLATFORM_NATIVE)
|
2004-07-13 14:01:41 +00:00
|
|
|
_ctype_,
|
2010-05-14 12:37:05 +00:00
|
|
|
#endif
|
2004-07-27 19:16:58 +00:00
|
|
|
atoi,
|
2005-02-22 12:19:12 +00:00
|
|
|
strchr,
|
|
|
|
|
strcat,
|
2010-02-14 06:26:16 +00:00
|
|
|
strlcat,
|
2006-09-16 16:18:11 +00:00
|
|
|
memchr,
|
2005-02-22 12:19:12 +00:00
|
|
|
memcmp,
|
2005-06-18 01:25:47 +00:00
|
|
|
strcasestr,
|
2006-10-31 11:17:00 +00:00
|
|
|
strtok_r,
|
2006-01-15 18:20:18 +00:00
|
|
|
/* unicode stuff */
|
|
|
|
|
utf8decode,
|
|
|
|
|
iso_decode,
|
|
|
|
|
utf16LEdecode,
|
|
|
|
|
utf16BEdecode,
|
|
|
|
|
utf8encode,
|
|
|
|
|
utf8length,
|
2006-09-16 16:18:11 +00:00
|
|
|
utf8seek,
|
2019-02-04 01:12:50 +00:00
|
|
|
|
2011-10-17 18:28:35 +00:00
|
|
|
/* the buflib memory management library */
|
|
|
|
|
buflib_init,
|
|
|
|
|
buflib_available,
|
|
|
|
|
buflib_alloc,
|
|
|
|
|
buflib_alloc_ex,
|
|
|
|
|
buflib_alloc_maximum,
|
|
|
|
|
buflib_buffer_in,
|
|
|
|
|
buflib_buffer_out,
|
|
|
|
|
buflib_free,
|
|
|
|
|
buflib_shrink,
|
|
|
|
|
buflib_get_data,
|
2003-06-29 16:33:04 +00:00
|
|
|
|
|
|
|
|
/* sound */
|
2023-03-21 02:15:33 +00:00
|
|
|
adjust_volume,
|
2005-04-01 13:41:03 +00:00
|
|
|
sound_set,
|
2019-07-20 09:46:14 +00:00
|
|
|
sound_current, /*stub*/
|
2008-03-24 00:35:53 +00:00
|
|
|
sound_default,
|
2006-01-15 18:20:18 +00:00
|
|
|
sound_min,
|
|
|
|
|
sound_max,
|
2008-03-24 00:35:53 +00:00
|
|
|
sound_unit,
|
|
|
|
|
sound_val2phys,
|
2010-05-24 16:42:32 +00:00
|
|
|
#ifdef AUDIOHW_HAVE_EQ
|
|
|
|
|
sound_enum_hw_eq_band_setting,
|
|
|
|
|
#endif
|
2020-07-17 04:01:32 +00:00
|
|
|
#if defined (HAVE_PITCHCONTROL)
|
2021-08-20 02:20:56 +00:00
|
|
|
sound_get_pitch,
|
2019-07-20 09:46:14 +00:00
|
|
|
sound_set_pitch,
|
2005-06-06 00:34:07 +00:00
|
|
|
#endif
|
2007-04-10 14:18:30 +00:00
|
|
|
&audio_master_sampr_list[0],
|
|
|
|
|
&hw_freq_sampr[0],
|
|
|
|
|
pcm_apply_settings,
|
2006-05-01 05:45:18 +00:00
|
|
|
pcm_play_data,
|
2005-06-18 01:25:47 +00:00
|
|
|
pcm_play_stop,
|
|
|
|
|
pcm_set_frequency,
|
|
|
|
|
pcm_is_playing,
|
2008-03-24 00:35:53 +00:00
|
|
|
pcm_play_lock,
|
|
|
|
|
pcm_play_unlock,
|
2011-06-29 06:37:04 +00:00
|
|
|
beep_play,
|
2007-04-10 14:18:30 +00:00
|
|
|
#ifdef HAVE_RECORDING
|
|
|
|
|
&rec_freq_sampr[0],
|
|
|
|
|
pcm_init_recording,
|
|
|
|
|
pcm_close_recording,
|
|
|
|
|
pcm_record_data,
|
|
|
|
|
pcm_stop_recording,
|
|
|
|
|
pcm_calculate_rec_peaks,
|
|
|
|
|
audio_set_recording_gain,
|
|
|
|
|
#endif /* HAVE_RECORDING */
|
2007-06-08 23:42:04 +00:00
|
|
|
#if INPUT_SRC_CAPS != 0
|
|
|
|
|
audio_set_output_source,
|
|
|
|
|
audio_set_input_source,
|
2003-06-29 16:33:04 +00:00
|
|
|
#endif
|
2012-05-01 07:58:27 +00:00
|
|
|
dsp_set_crossfeed_type ,
|
2012-03-27 23:52:15 +00:00
|
|
|
dsp_eq_enable,
|
2008-03-24 00:35:53 +00:00
|
|
|
dsp_dither_enable,
|
2012-05-08 14:34:26 +00:00
|
|
|
#ifdef HAVE_PITCHCONTROL
|
2021-08-20 02:20:56 +00:00
|
|
|
dsp_get_timestretch,
|
2012-03-27 23:52:15 +00:00
|
|
|
dsp_set_timestretch,
|
2021-08-20 02:20:56 +00:00
|
|
|
dsp_timestretch_enable,
|
|
|
|
|
dsp_timestretch_available,
|
2012-03-27 23:52:15 +00:00
|
|
|
#endif
|
2008-03-24 00:35:53 +00:00
|
|
|
dsp_configure,
|
2012-03-27 23:52:15 +00:00
|
|
|
dsp_get_config,
|
2008-03-24 00:35:53 +00:00
|
|
|
dsp_process,
|
2005-06-18 01:25:47 +00:00
|
|
|
|
2011-08-03 09:52:23 +00:00
|
|
|
mixer_channel_status,
|
|
|
|
|
mixer_channel_get_buffer,
|
|
|
|
|
mixer_channel_calculate_peaks,
|
|
|
|
|
mixer_channel_play_data,
|
|
|
|
|
mixer_channel_play_pause,
|
|
|
|
|
mixer_channel_stop,
|
|
|
|
|
mixer_channel_set_amplitude,
|
|
|
|
|
mixer_channel_get_bytes_waiting,
|
2012-12-09 20:04:17 +00:00
|
|
|
mixer_channel_set_buffer_hook,
|
2013-12-14 22:15:45 +00:00
|
|
|
mixer_set_frequency,
|
|
|
|
|
mixer_get_frequency,
|
2011-08-03 09:52:23 +00:00
|
|
|
|
2019-02-04 01:12:50 +00:00
|
|
|
pcmbuf_fade,
|
2021-08-20 02:20:56 +00:00
|
|
|
pcmbuf_set_low_latency,
|
2011-08-03 09:52:23 +00:00
|
|
|
system_sound_play,
|
|
|
|
|
keyclick_click,
|
2019-07-20 09:46:14 +00:00
|
|
|
|
|
|
|
|
/* metadata */
|
|
|
|
|
get_metadata,
|
|
|
|
|
mp3info,
|
|
|
|
|
count_mp3_frames,
|
|
|
|
|
create_xing_header,
|
|
|
|
|
#ifdef HAVE_TAGCACHE
|
|
|
|
|
tagcache_search,
|
|
|
|
|
tagcache_search_set_uniqbuf,
|
|
|
|
|
tagcache_search_add_filter,
|
|
|
|
|
tagcache_get_next,
|
|
|
|
|
tagcache_retrieve,
|
|
|
|
|
tagcache_search_finish,
|
|
|
|
|
tagcache_get_numeric,
|
2022-11-30 14:50:18 +00:00
|
|
|
tagcache_get_stat,
|
|
|
|
|
#if defined(HAVE_TC_RAMCACHE)
|
|
|
|
|
tagcache_is_in_ram,
|
|
|
|
|
#if defined(HAVE_DIRCACHE)
|
2019-07-20 09:46:14 +00:00
|
|
|
tagcache_fill_tags,
|
|
|
|
|
#endif
|
2022-11-30 14:50:18 +00:00
|
|
|
#endif
|
2023-03-21 02:15:33 +00:00
|
|
|
tagtree_subentries_do_action,
|
2019-07-20 09:46:14 +00:00
|
|
|
#endif /* HAVE_TAGCACHE */
|
|
|
|
|
|
|
|
|
|
#ifdef HAVE_ALBUMART
|
|
|
|
|
search_albumart_files,
|
|
|
|
|
#endif
|
|
|
|
|
|
2004-07-27 14:10:48 +00:00
|
|
|
/* playback control */
|
2020-07-23 13:49:34 +00:00
|
|
|
playlist_get_current,
|
2007-04-10 14:18:30 +00:00
|
|
|
playlist_amount,
|
|
|
|
|
playlist_resume,
|
2012-12-09 20:04:17 +00:00
|
|
|
playlist_resume_track,
|
2023-03-21 02:15:33 +00:00
|
|
|
playlist_set_modified,
|
2007-04-10 14:18:30 +00:00
|
|
|
playlist_start,
|
2009-06-01 21:08:49 +00:00
|
|
|
playlist_add,
|
|
|
|
|
playlist_sync,
|
|
|
|
|
playlist_remove_all_tracks,
|
|
|
|
|
playlist_create,
|
|
|
|
|
playlist_insert_track,
|
2009-07-05 18:07:58 +00:00
|
|
|
playlist_insert_directory,
|
2022-09-19 10:10:54 +00:00
|
|
|
playlist_insert_playlist,
|
2009-06-01 21:08:49 +00:00
|
|
|
playlist_shuffle,
|
2022-09-19 10:10:54 +00:00
|
|
|
warn_on_pl_erase,
|
2010-09-01 21:29:34 +00:00
|
|
|
audio_play,
|
2005-04-04 12:06:29 +00:00
|
|
|
audio_stop,
|
|
|
|
|
audio_pause,
|
|
|
|
|
audio_resume,
|
|
|
|
|
audio_next,
|
|
|
|
|
audio_prev,
|
|
|
|
|
audio_ff_rewind,
|
|
|
|
|
audio_next_track,
|
|
|
|
|
audio_status,
|
|
|
|
|
audio_current_track,
|
|
|
|
|
audio_flush_and_reload_tracks,
|
|
|
|
|
audio_get_file_pos,
|
2021-08-20 02:20:56 +00:00
|
|
|
#ifdef PLUGIN_USE_IRAM
|
|
|
|
|
audio_hard_stop,
|
|
|
|
|
#endif
|
2004-07-27 19:16:58 +00:00
|
|
|
|
2006-01-15 18:20:18 +00:00
|
|
|
/* menu */
|
2019-02-04 01:12:50 +00:00
|
|
|
root_menu_get_options,
|
2007-03-07 13:00:46 +00:00
|
|
|
do_menu,
|
2019-02-04 01:12:50 +00:00
|
|
|
root_menu_set_default,
|
|
|
|
|
root_menu_write_to_cfg,
|
|
|
|
|
root_menu_load_from_cfg,
|
|
|
|
|
|
2007-05-08 11:55:43 +00:00
|
|
|
/* options */
|
2008-12-10 08:57:10 +00:00
|
|
|
get_settings_list,
|
2007-07-11 05:41:23 +00:00
|
|
|
find_setting,
|
2019-02-04 01:12:50 +00:00
|
|
|
settings_save,
|
2007-07-11 05:41:23 +00:00
|
|
|
option_screen,
|
2006-03-21 09:36:13 +00:00
|
|
|
set_option,
|
2007-06-22 09:34:57 +00:00
|
|
|
set_bool_options,
|
2006-09-16 16:18:11 +00:00
|
|
|
set_int,
|
2019-02-04 01:12:50 +00:00
|
|
|
set_int_ex,
|
2006-09-16 16:18:11 +00:00
|
|
|
set_bool,
|
2007-06-22 09:34:57 +00:00
|
|
|
#ifdef HAVE_LCD_COLOR
|
|
|
|
|
set_color,
|
|
|
|
|
#endif
|
2006-01-15 18:20:18 +00:00
|
|
|
|
2006-02-07 20:38:55 +00:00
|
|
|
/* power */
|
|
|
|
|
battery_level,
|
|
|
|
|
battery_level_safe,
|
|
|
|
|
battery_time,
|
2006-05-01 05:45:18 +00:00
|
|
|
battery_voltage,
|
2022-09-19 10:10:54 +00:00
|
|
|
battery_current,
|
2007-02-18 05:32:06 +00:00
|
|
|
#if CONFIG_CHARGING
|
2006-02-07 20:38:55 +00:00
|
|
|
charger_inserted,
|
2009-03-24 21:59:44 +00:00
|
|
|
# if CONFIG_CHARGING >= CHARGING_MONITOR
|
2006-02-07 20:38:55 +00:00
|
|
|
charging_state,
|
|
|
|
|
# endif
|
|
|
|
|
#endif
|
2019-07-20 09:46:14 +00:00
|
|
|
/* usb */
|
2015-01-02 21:15:31 +00:00
|
|
|
usb_inserted,
|
2019-07-20 09:46:14 +00:00
|
|
|
usb_acknowledge,
|
|
|
|
|
#ifdef USB_ENABLE_HID
|
|
|
|
|
usb_hid_send,
|
|
|
|
|
#endif
|
2003-06-29 16:33:04 +00:00
|
|
|
/* misc */
|
2010-06-21 16:53:00 +00:00
|
|
|
#if (CONFIG_PLATFORM & PLATFORM_NATIVE)
|
2014-04-28 13:15:06 +00:00
|
|
|
__errno,
|
2010-02-14 06:26:16 +00:00
|
|
|
#endif
|
2021-08-20 02:20:56 +00:00
|
|
|
led,
|
2003-06-29 16:33:04 +00:00
|
|
|
srand,
|
|
|
|
|
rand,
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
(void *)qsort,
|
2003-07-09 22:36:23 +00:00
|
|
|
kbd_input,
|
2003-07-25 23:04:59 +00:00
|
|
|
get_time,
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
gmtime_r,
|
2007-07-30 05:19:05 +00:00
|
|
|
#if CONFIG_RTC
|
|
|
|
|
mktime,
|
|
|
|
|
#endif
|
2019-07-20 09:46:14 +00:00
|
|
|
|
2004-06-10 13:29:52 +00:00
|
|
|
#if defined(DEBUG) || defined(SIMULATOR)
|
|
|
|
|
debugf,
|
2005-07-26 20:01:11 +00:00
|
|
|
#endif
|
|
|
|
|
#ifdef ROCKBOX_HAS_LOGF
|
2006-11-11 08:00:28 +00:00
|
|
|
_logf,
|
2004-06-10 13:29:52 +00:00
|
|
|
#endif
|
2009-01-05 10:31:19 +00:00
|
|
|
codec_thread_do_callback,
|
2007-10-16 01:25:17 +00:00
|
|
|
codec_load_file,
|
2011-04-27 03:08:23 +00:00
|
|
|
codec_run_proc,
|
2011-02-23 14:31:13 +00:00
|
|
|
codec_close,
|
2007-10-16 01:25:17 +00:00
|
|
|
get_codec_filename,
|
2010-05-24 16:42:32 +00:00
|
|
|
find_array_ptr,
|
|
|
|
|
remove_array_ptr,
|
|
|
|
|
round_value_to_list32,
|
2019-07-20 09:46:14 +00:00
|
|
|
|
2005-04-25 07:42:10 +00:00
|
|
|
read_bmp_file,
|
2009-05-06 04:53:56 +00:00
|
|
|
read_bmp_fd,
|
|
|
|
|
#ifdef HAVE_JPEG
|
|
|
|
|
read_jpeg_file,
|
|
|
|
|
read_jpeg_fd,
|
|
|
|
|
#endif
|
2006-01-15 18:20:18 +00:00
|
|
|
screen_dump_set_hook,
|
2005-06-18 01:25:47 +00:00
|
|
|
|
2006-09-26 19:25:52 +00:00
|
|
|
#ifdef HAVE_WHEEL_POSITION
|
|
|
|
|
wheel_status,
|
|
|
|
|
wheel_send_events,
|
|
|
|
|
#endif
|
2006-10-31 11:17:00 +00:00
|
|
|
|
2020-10-25 23:34:45 +00:00
|
|
|
#if defined(IRIVER_H100_SERIES) || defined(IRIVER_H300_SERIES)
|
2007-01-12 18:34:00 +00:00
|
|
|
/* Routines for the iriver_flash -plugin. */
|
|
|
|
|
detect_original_firmware,
|
|
|
|
|
detect_flashed_ramimage,
|
|
|
|
|
detect_flashed_romimage,
|
2007-10-16 01:25:17 +00:00
|
|
|
#endif
|
2007-12-29 19:46:35 +00:00
|
|
|
|
2019-07-20 09:46:14 +00:00
|
|
|
/*plugin*/
|
2020-09-20 15:24:58 +00:00
|
|
|
plugin_open,
|
2019-07-20 09:46:14 +00:00
|
|
|
plugin_get_buffer,
|
|
|
|
|
plugin_get_audio_buffer, /* defined in plugin.c */
|
|
|
|
|
plugin_release_audio_buffer, /* defined in plugin.c */
|
|
|
|
|
plugin_tsr, /* defined in plugin.c */
|
|
|
|
|
plugin_get_current_filename,
|
2022-03-25 14:28:35 +00:00
|
|
|
plugin_reserve_buffer,
|
2022-09-19 10:10:54 +00:00
|
|
|
|
|
|
|
|
/* reboot and poweroff */
|
|
|
|
|
sys_poweroff,
|
|
|
|
|
sys_reboot,
|
|
|
|
|
|
|
|
|
|
/* pathfuncs */
|
2023-03-21 02:15:33 +00:00
|
|
|
fix_path_part,
|
2022-04-16 14:39:56 +00:00
|
|
|
#ifdef HAVE_MULTIVOLUME
|
2022-04-16 11:13:19 +00:00
|
|
|
path_strip_volume,
|
2022-04-16 14:39:56 +00:00
|
|
|
#endif
|
2022-09-19 10:10:54 +00:00
|
|
|
|
|
|
|
|
/* new stuff at the end, sort into place next time
|
|
|
|
|
the API gets incompatible */
|
2022-05-25 18:11:29 +00:00
|
|
|
|
2003-06-29 16:33:04 +00:00
|
|
|
};
|
|
|
|
|
|
2013-05-30 09:24:16 +00:00
|
|
|
static int plugin_buffer_handle;
|
2017-10-14 09:59:09 +00:00
|
|
|
static size_t plugin_buffer_size;
|
2013-05-30 09:24:16 +00:00
|
|
|
|
2008-05-13 09:57:56 +00:00
|
|
|
int plugin_load(const char* plugin, const void* parameter)
|
2003-06-29 16:33:04 +00:00
|
|
|
{
|
2010-09-09 16:17:21 +00:00
|
|
|
struct plugin_header *p_hdr;
|
|
|
|
|
struct lc_header *hdr;
|
2023-03-21 02:15:33 +00:00
|
|
|
const char * resume_plugin = NULL;
|
|
|
|
|
|
|
|
|
|
if (!plugin)
|
|
|
|
|
return PLUGIN_ERROR;
|
2006-11-15 06:14:27 +00:00
|
|
|
|
2010-08-27 00:29:50 +00:00
|
|
|
if (current_plugin_handle && pfn_tsr_exit)
|
|
|
|
|
{ /* if we have a resident old plugin and a callback */
|
2023-03-21 02:15:33 +00:00
|
|
|
bool reenter = (strcmp(current_plugin, plugin) == 0);
|
|
|
|
|
int exit_status = pfn_tsr_exit(reenter);
|
|
|
|
|
if (exit_status == PLUGIN_TSR_CONTINUE)
|
2006-10-31 11:17:00 +00:00
|
|
|
{
|
|
|
|
|
/* not allowing another plugin to load */
|
|
|
|
|
return PLUGIN_OK;
|
|
|
|
|
}
|
2023-03-21 02:15:33 +00:00
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
lc_close(current_plugin_handle);
|
|
|
|
|
current_plugin_handle = pfn_tsr_exit = NULL;
|
|
|
|
|
plugin_buffer_handle = core_free(plugin_buffer_handle);
|
|
|
|
|
|
|
|
|
|
if (!reenter)
|
|
|
|
|
resume_plugin = strdupa(current_plugin);
|
|
|
|
|
else if (exit_status == PLUGIN_TSR_TERMINATE)
|
|
|
|
|
return PLUGIN_OK; /* don't even load the new plugin either */
|
|
|
|
|
}
|
2004-04-30 20:37:11 +00:00
|
|
|
}
|
2007-04-10 14:18:30 +00:00
|
|
|
|
2022-10-10 01:28:40 +00:00
|
|
|
#ifdef HAVE_DISK_STORAGE
|
|
|
|
|
if (!storage_disk_is_active())
|
|
|
|
|
splash(0, ID2P(LANG_WAIT));
|
|
|
|
|
#endif
|
2007-08-08 07:27:07 +00:00
|
|
|
strcpy(current_plugin, plugin);
|
2010-08-27 00:29:50 +00:00
|
|
|
current_plugin_handle = lc_open(plugin, pluginbuf, PLUGIN_BUFFER_SIZE);
|
|
|
|
|
if (current_plugin_handle == NULL) {
|
2008-08-15 08:27:39 +00:00
|
|
|
splashf(HZ*2, str(LANG_PLUGIN_CANT_OPEN), plugin);
|
2003-06-29 16:33:04 +00:00
|
|
|
return -1;
|
2006-01-16 23:20:58 +00:00
|
|
|
}
|
2006-01-17 20:39:44 +00:00
|
|
|
|
2010-09-09 16:17:21 +00:00
|
|
|
p_hdr = lc_get_header(current_plugin_handle);
|
|
|
|
|
hdr = p_hdr ? &p_hdr->lc_hdr : NULL;
|
2019-02-04 01:12:50 +00:00
|
|
|
|
2010-08-27 00:29:50 +00:00
|
|
|
if (hdr == NULL
|
2006-01-18 08:34:32 +00:00
|
|
|
|| hdr->magic != PLUGIN_MAGIC
|
2006-01-17 20:39:44 +00:00
|
|
|
|| hdr->target_id != TARGET_ID
|
2010-08-27 00:29:50 +00:00
|
|
|
#if (CONFIG_PLATFORM & PLATFORM_NATIVE)
|
2006-01-17 20:39:44 +00:00
|
|
|
|| hdr->load_addr != pluginbuf
|
2010-08-27 00:29:50 +00:00
|
|
|
|| hdr->end_addr > pluginbuf + PLUGIN_BUFFER_SIZE
|
|
|
|
|
#endif
|
|
|
|
|
)
|
|
|
|
|
{
|
|
|
|
|
lc_close(current_plugin_handle);
|
2019-02-04 01:12:50 +00:00
|
|
|
splash(HZ*2, ID2P(LANG_PLUGIN_WRONG_MODEL));
|
2006-01-15 18:20:18 +00:00
|
|
|
return -1;
|
|
|
|
|
}
|
2023-03-14 12:19:48 +00:00
|
|
|
|
|
|
|
|
if (hdr->api_version != PLUGIN_API_VERSION ||
|
|
|
|
|
p_hdr->api_size > sizeof(struct plugin_api))
|
2010-08-27 00:29:50 +00:00
|
|
|
{
|
|
|
|
|
lc_close(current_plugin_handle);
|
2019-02-04 01:12:50 +00:00
|
|
|
splash(HZ*2, ID2P(LANG_PLUGIN_WRONG_VERSION));
|
2006-01-15 18:20:18 +00:00
|
|
|
return -1;
|
|
|
|
|
}
|
2023-03-14 12:19:48 +00:00
|
|
|
|
2010-08-27 00:29:50 +00:00
|
|
|
#if (CONFIG_PLATFORM & PLATFORM_NATIVE)
|
2019-07-17 05:39:14 +00:00
|
|
|
/* tlsf crashes observed on arm with 0x4 aligned addresses */
|
|
|
|
|
plugin_size = ALIGN_UP(hdr->end_addr - pluginbuf, 0x8);
|
2010-08-27 00:29:50 +00:00
|
|
|
#else
|
|
|
|
|
plugin_size = 0;
|
2003-06-29 16:33:04 +00:00
|
|
|
#endif
|
|
|
|
|
|
2010-09-09 16:17:21 +00:00
|
|
|
*(p_hdr->api) = &rockbox_api;
|
2020-10-07 06:01:35 +00:00
|
|
|
lcd_set_viewport(NULL);
|
2006-02-16 22:36:04 +00:00
|
|
|
lcd_clear_display();
|
2006-02-14 00:39:11 +00:00
|
|
|
|
2006-02-18 00:13:44 +00:00
|
|
|
#ifdef HAVE_REMOTE_LCD
|
2020-10-07 06:01:35 +00:00
|
|
|
lcd_remote_set_viewport(NULL);
|
2006-02-18 00:13:44 +00:00
|
|
|
lcd_remote_clear_display();
|
|
|
|
|
lcd_remote_update();
|
|
|
|
|
#endif
|
2022-12-12 02:49:06 +00:00
|
|
|
if (get_current_activity() == ACTIVITY_WPS)
|
|
|
|
|
push_activity_without_refresh(ACTIVITY_PLUGIN);
|
|
|
|
|
else
|
|
|
|
|
push_current_activity(ACTIVITY_PLUGIN);
|
2011-08-30 14:01:45 +00:00
|
|
|
/* some plugins assume the entry cache doesn't move and save pointers to it
|
|
|
|
|
* they should be fixed properly instead of this lock */
|
|
|
|
|
tree_lock_cache(tree_get_context());
|
2006-02-18 00:13:44 +00:00
|
|
|
|
2009-12-09 07:25:46 +00:00
|
|
|
FOR_NB_SCREENS(i)
|
|
|
|
|
viewportmanager_theme_enable(i, false, NULL);
|
2019-02-04 01:12:50 +00:00
|
|
|
|
2009-04-19 12:46:04 +00:00
|
|
|
#ifdef HAVE_TOUCHSCREEN
|
|
|
|
|
touchscreen_set_mode(TOUCHSCREEN_BUTTON);
|
|
|
|
|
#endif
|
2007-03-21 07:26:07 +00:00
|
|
|
|
2013-05-30 09:24:16 +00:00
|
|
|
/* allow voice to back off if the plugin needs lots of memory */
|
2019-02-04 01:12:50 +00:00
|
|
|
if (!global_settings.talk_menu)
|
|
|
|
|
talk_buffer_set_policy(TALK_BUFFER_LOOSE);
|
2013-05-30 09:24:16 +00:00
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
plugin_check_open_close__enter();
|
2010-02-11 19:15:37 +00:00
|
|
|
|
2011-10-15 19:35:02 +00:00
|
|
|
int rc = p_hdr->entry_point(parameter);
|
2019-02-04 01:12:50 +00:00
|
|
|
|
2011-08-30 14:01:45 +00:00
|
|
|
tree_unlock_cache(tree_get_context());
|
2022-12-12 02:49:06 +00:00
|
|
|
|
2022-12-14 08:06:04 +00:00
|
|
|
pop_current_activity_without_refresh();
|
2022-12-16 12:04:56 +00:00
|
|
|
if (get_current_activity() != ACTIVITY_WPS)
|
2022-12-12 02:49:06 +00:00
|
|
|
{
|
2022-12-16 07:02:46 +00:00
|
|
|
FOR_NB_SCREENS(i)
|
2022-12-14 08:06:04 +00:00
|
|
|
skin_update(CUSTOM_STATUSBAR, i, SKIN_REFRESH_ALL);
|
2022-12-12 02:49:06 +00:00
|
|
|
}
|
2010-02-11 19:15:37 +00:00
|
|
|
|
2010-08-27 00:29:50 +00:00
|
|
|
if (!pfn_tsr_exit)
|
|
|
|
|
{ /* close handle if plugin is no tsr one */
|
|
|
|
|
lc_close(current_plugin_handle);
|
|
|
|
|
current_plugin_handle = NULL;
|
2022-10-15 13:08:09 +00:00
|
|
|
plugin_buffer_handle = core_free(plugin_buffer_handle);
|
2010-08-27 00:29:50 +00:00
|
|
|
}
|
|
|
|
|
|
2013-05-30 09:24:16 +00:00
|
|
|
talk_buffer_set_policy(TALK_BUFFER_DEFAULT);
|
|
|
|
|
|
2009-08-04 05:22:10 +00:00
|
|
|
/* Go back to the global setting in case the plugin changed it */
|
2009-04-19 12:46:04 +00:00
|
|
|
#ifdef HAVE_TOUCHSCREEN
|
2009-08-04 05:22:10 +00:00
|
|
|
touchscreen_set_mode(global_settings.touch_mode);
|
2009-04-19 12:46:04 +00:00
|
|
|
#endif
|
2020-10-07 06:01:35 +00:00
|
|
|
/* restore default vp */
|
|
|
|
|
lcd_set_viewport(NULL);
|
2011-09-24 13:19:34 +00:00
|
|
|
screen_helper_setfont(FONT_UI);
|
2005-07-14 18:12:27 +00:00
|
|
|
#if LCD_DEPTH > 1
|
2006-11-13 00:45:21 +00:00
|
|
|
#ifdef HAVE_LCD_COLOR
|
2006-05-01 05:45:18 +00:00
|
|
|
lcd_set_drawinfo(DRMODE_SOLID, global_settings.fg_color,
|
2006-02-26 02:48:05 +00:00
|
|
|
global_settings.bg_color);
|
|
|
|
|
#else
|
2005-11-16 21:09:23 +00:00
|
|
|
lcd_set_drawinfo(DRMODE_SOLID, LCD_DEFAULT_FG, LCD_DEFAULT_BG);
|
2006-02-26 02:48:05 +00:00
|
|
|
#endif
|
2005-07-19 10:33:41 +00:00
|
|
|
#else /* LCD_DEPTH == 1 */
|
2005-06-24 22:33:21 +00:00
|
|
|
lcd_set_drawmode(DRMODE_SOLID);
|
2005-07-19 10:33:41 +00:00
|
|
|
#endif /* LCD_DEPTH */
|
2006-02-18 00:13:44 +00:00
|
|
|
|
|
|
|
|
#ifdef HAVE_REMOTE_LCD
|
2020-10-07 06:01:35 +00:00
|
|
|
lcd_remote_set_viewport(NULL);
|
|
|
|
|
|
2006-10-09 07:00:50 +00:00
|
|
|
#if LCD_REMOTE_DEPTH > 1
|
2006-10-14 01:32:58 +00:00
|
|
|
lcd_remote_set_drawinfo(DRMODE_SOLID, LCD_REMOTE_DEFAULT_FG,
|
2006-10-09 07:00:50 +00:00
|
|
|
LCD_REMOTE_DEFAULT_BG);
|
|
|
|
|
#else
|
2006-02-18 00:13:44 +00:00
|
|
|
lcd_remote_set_drawmode(DRMODE_SOLID);
|
2006-10-09 07:00:50 +00:00
|
|
|
#endif
|
2009-07-05 22:12:42 +00:00
|
|
|
#endif
|
2007-10-16 01:25:17 +00:00
|
|
|
|
2009-12-10 13:53:38 +00:00
|
|
|
lcd_clear_display();
|
2010-07-24 10:53:58 +00:00
|
|
|
#ifdef HAVE_REMOTE_LCD
|
2009-12-10 13:53:38 +00:00
|
|
|
lcd_remote_clear_display();
|
|
|
|
|
#endif
|
|
|
|
|
|
2009-12-09 07:25:46 +00:00
|
|
|
FOR_NB_SCREENS(i)
|
2011-09-14 12:47:26 +00:00
|
|
|
viewportmanager_theme_undo(i, true);
|
2007-10-16 01:25:17 +00:00
|
|
|
|
Rewrite filesystem code (WIP)
This patch redoes the filesystem code from the FAT driver up to the
clipboard code in onplay.c.
Not every aspect of this is finished therefore it is still "WIP". I
don't wish to do too much at once (haha!). What is left to do is get
dircache back in the sim and find an implementation for the dircache
indicies in the tagcache and playlist code or do something else that
has the same benefit. Leaving these out for now does not make anything
unusable. All the basics are done.
Phone app code should probably get vetted (and app path handling
just plain rewritten as environment expansions); the SDL app and
Android run well.
Main things addressed:
1) Thread safety: There is none right now in the trunk code. Most of
what currently works is luck when multiple threads are involved or
multiple descriptors to the same file are open.
2) POSIX compliance: Many of the functions behave nothing like their
counterparts on a host system. This leads to inconsistent code or very
different behavior from native to hosted. One huge offender was
rename(). Going point by point would fill a book.
3) Actual running RAM usage: Many targets will use less RAM and less
stack space (some more RAM because I upped the number of cache buffers
for large memory). There's very little memory lying fallow in rarely-used
areas (see 'Key core changes' below). Also, all targets may open the same
number of directory streams whereas before those with less than 8MB RAM
were limited to 8, not 12 implying those targets will save slightly
less.
4) Performance: The test_disk plugin shows markedly improved performance,
particularly in the area of (uncached) directory scanning, due partly to
more optimal directory reading and to a better sector cache algorithm.
Uncached times tend to be better while there is a bit of a slowdown in
dircache due to it being a bit heavier of an implementation. It's not
noticeable by a human as far as I can say.
Key core changes:
1) Files and directories share core code and data structures.
2) The filesystem code knows which descriptors refer to same file.
This ensures that changes from one stream are appropriately reflected
in every open descriptor for that file (fileobj_mgr.c).
3) File and directory cache buffers are borrowed from the main sector
cache. This means that when they are not in use by a file, they are not
wasted, but used for the cache. Most of the time, only a few of them
are needed. It also means that adding more file and directory handles
is less expensive. All one must do in ensure a large enough cache to
borrow from.
4) Relative path components are supported and the namespace is unified.
It does not support full relative paths to an implied current directory;
what is does support is use of "." and "..". Adding the former would
not be very difficult. The namespace is unified in the sense that
volumes may be specified several times along with relative parts, e.g.:
"/<0>/foo/../../<1>/bar" :<=> "/<1>/bar".
5) Stack usage is down due to sharing of data, static allocation and
less duplication of strings on the stack. This requires more
serialization than I would like but since the number of threads is
limited to a low number, the tradoff in favor of the stack seems
reasonable.
6) Separates and heirarchicalizes (sic) the SIM and APP filesystem
code. SIM path and volume handling is just like the target. Some
aspects of the APP file code get more straightforward (e.g. no path
hashing is needed).
Dircache:
Deserves its own section. Dircache is new but pays homage to the old.
The old one was not compatible and so it, since it got redone, does
all the stuff it always should have done such as:
1) It may be update and used at any time during the build process.
No longer has one to wait for it to finish building to do basic file
management (create, remove, rename, etc.).
2) It does not need to be either fully scanned or completely disabled;
it can be incomplete (i.e. overfilled, missing paths), still be
of benefit and be correct.
3) Handles mounting and dismounting of individual volumes which means
a full rebuild is not needed just because you pop a new SD card in the
slot. Now, because it reuses its freed entry data, may rebuild only
that volume.
4) Much more fundamental to the file code. When it is built, it is
the keeper of the master file list whether enabled or not ("disabled"
is just a state of the cache). Its must always to ready to be started
and bind all streams opened prior to being enabled.
5) Maintains any short filenames in OEM format which means that it does
not need to be rebuilt when changing the default codepage.
Miscellaneous Compatibility:
1) Update any other code that would otherwise not work such as the
hotswap mounting code in various card drivers.
2) File management: Clipboard needed updating because of the behavioral
changes. Still needs a little more work on some finer points.
3) Remove now-obsolete functionality such as the mutex's "no preempt"
flag (which was only for the prior FAT driver).
4) struct dirinfo uses time_t rather than raw FAT directory entry
time fields. I plan to follow up on genericizing everything there
(i.e. no FAT attributes).
5) unicode.c needed some redoing so that the file code does not try
try to load codepages during a scan, which is actually a problem with
the current code. The default codepage, if any is required, is now
kept in RAM separarately (bufalloced) from codepages specified to
iso_decode() (which must not be bufalloced because the conversion
may be done by playback threads).
Brings with it some additional reusable core code:
1) Revised file functions: Reusable code that does things such as
safe path concatenation and parsing without buffer limitations or
data duplication. Variants that copy or alter the input path may be
based off these.
To do:
1) Put dircache functionality back in the sim. Treating it internally
as a different kind of file system seems the best approach at this
time.
2) Restore use of dircache indexes in the playlist and database or
something effectively the same. Since the cache doesn't have to be
complete in order to be used, not getting a hit on the cache doesn't
unambiguously say if the path exists or not.
Change-Id: Ia30f3082a136253e3a0eae0784e3091d138915c8
Reviewed-on: http://gerrit.rockbox.org/566
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
2013-08-06 02:02:45 +00:00
|
|
|
plugin_check_open_close__exit();
|
2010-02-11 19:15:37 +00:00
|
|
|
|
2019-08-06 02:05:32 +00:00
|
|
|
status_save();
|
|
|
|
|
|
2009-07-05 22:12:42 +00:00
|
|
|
if (rc == PLUGIN_ERROR)
|
|
|
|
|
splash(HZ*2, str(LANG_PLUGIN_ERROR));
|
2008-12-13 02:48:36 +00:00
|
|
|
|
2023-03-21 02:15:33 +00:00
|
|
|
if (resume_plugin && rc != PLUGIN_GOTO_PLUGIN && !pfn_tsr_exit)
|
|
|
|
|
{
|
|
|
|
|
/*plugin = resume_plugin;*/
|
|
|
|
|
/*parameter = rockbox_api.plugin_tsr;*/
|
|
|
|
|
return plugin_load(resume_plugin, rockbox_api.plugin_tsr);
|
|
|
|
|
}
|
2009-07-05 22:12:42 +00:00
|
|
|
return rc;
|
2003-06-29 16:33:04 +00:00
|
|
|
}
|
|
|
|
|
|
2022-03-25 14:28:35 +00:00
|
|
|
/* For Terminate Stay Resident plugins
|
|
|
|
|
* Locks buffer_size bytes of the plugin buffer
|
|
|
|
|
* freed on plugin exit; call plugin_get_buffer first then reserve all
|
|
|
|
|
* or a portion with plugin_reserve_buffer()
|
|
|
|
|
* Returns size of buffer remaining */
|
|
|
|
|
size_t plugin_reserve_buffer(size_t buffer_size)
|
|
|
|
|
{
|
|
|
|
|
size_t locked_size = 0;
|
|
|
|
|
|
|
|
|
|
if (current_plugin_handle)
|
|
|
|
|
{
|
|
|
|
|
locked_size = ALIGN_UP(plugin_size + buffer_size, 0x8);
|
|
|
|
|
if (locked_size > PLUGIN_BUFFER_SIZE)
|
|
|
|
|
locked_size = PLUGIN_BUFFER_SIZE;
|
|
|
|
|
|
|
|
|
|
plugin_size = locked_size;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (PLUGIN_BUFFER_SIZE - locked_size);
|
|
|
|
|
}
|
|
|
|
|
|
2003-07-25 23:04:59 +00:00
|
|
|
/* Returns a pointer to the portion of the plugin buffer that is not already
|
|
|
|
|
being used. If no plugin is loaded, returns the entire plugin buffer */
|
2007-04-21 18:38:25 +00:00
|
|
|
void* plugin_get_buffer(size_t *buffer_size)
|
2003-07-25 23:04:59 +00:00
|
|
|
{
|
|
|
|
|
int buffer_pos;
|
|
|
|
|
|
2010-08-27 00:29:50 +00:00
|
|
|
if (current_plugin_handle)
|
2003-07-25 23:04:59 +00:00
|
|
|
{
|
|
|
|
|
if (plugin_size >= PLUGIN_BUFFER_SIZE)
|
|
|
|
|
return NULL;
|
2006-05-01 05:45:18 +00:00
|
|
|
|
2003-07-25 23:04:59 +00:00
|
|
|
*buffer_size = PLUGIN_BUFFER_SIZE-plugin_size;
|
|
|
|
|
buffer_pos = plugin_size;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
*buffer_size = PLUGIN_BUFFER_SIZE;
|
|
|
|
|
buffer_pos = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return &pluginbuf[buffer_pos];
|
|
|
|
|
}
|
|
|
|
|
|
2006-05-01 05:45:18 +00:00
|
|
|
/* Returns a pointer to the mp3 buffer.
|
2006-11-06 18:07:30 +00:00
|
|
|
Playback gets stopped, to avoid conflicts.
|
|
|
|
|
Talk buffer is stolen as well.
|
|
|
|
|
*/
|
2014-04-09 07:00:36 +00:00
|
|
|
static void* plugin_get_audio_buffer(size_t *buffer_size)
|
2004-01-30 22:30:40 +00:00
|
|
|
{
|
2017-10-14 09:59:09 +00:00
|
|
|
if (plugin_buffer_handle <= 0)
|
|
|
|
|
{
|
2022-10-15 22:55:39 +00:00
|
|
|
plugin_buffer_handle = core_alloc_maximum(&plugin_buffer_size,
|
2022-01-18 18:57:06 +00:00
|
|
|
&buflib_ops_locked);
|
2017-10-14 09:59:09 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (buffer_size)
|
|
|
|
|
*buffer_size = plugin_buffer_size;
|
|
|
|
|
|
2013-05-30 09:24:16 +00:00
|
|
|
return core_get_data(plugin_buffer_handle);
|
2006-11-18 02:18:29 +00:00
|
|
|
}
|
|
|
|
|
|
2014-04-09 07:00:36 +00:00
|
|
|
static void plugin_release_audio_buffer(void)
|
|
|
|
|
{
|
|
|
|
|
if (plugin_buffer_handle > 0)
|
2017-10-14 09:59:09 +00:00
|
|
|
{
|
2014-04-09 07:00:36 +00:00
|
|
|
plugin_buffer_handle = core_free(plugin_buffer_handle);
|
2017-10-14 09:59:09 +00:00
|
|
|
plugin_buffer_size = 0;
|
|
|
|
|
}
|
2014-04-09 07:00:36 +00:00
|
|
|
}
|
|
|
|
|
|
2004-04-30 20:37:11 +00:00
|
|
|
/* The plugin wants to stay resident after leaving its main function, e.g.
|
2006-05-01 05:45:18 +00:00
|
|
|
runs from timer or own thread. The callback is registered to later
|
2004-04-30 20:37:11 +00:00
|
|
|
instruct it to free its resources before a new plugin gets loaded. */
|
2023-03-21 02:15:33 +00:00
|
|
|
static void plugin_tsr(int (*exit_callback)(bool))
|
2004-04-30 20:37:11 +00:00
|
|
|
{
|
|
|
|
|
pfn_tsr_exit = exit_callback; /* remember the callback for later */
|
|
|
|
|
}
|
2007-08-06 13:42:52 +00:00
|
|
|
|
2021-10-23 03:17:33 +00:00
|
|
|
int plugin_open(const char *plugin, const char *parameter)
|
2020-08-01 02:45:10 +00:00
|
|
|
{
|
|
|
|
|
open_plugin_add_path(ID2P(LANG_OPEN_PLUGIN), plugin, parameter);
|
|
|
|
|
return PLUGIN_GOTO_PLUGIN;
|
|
|
|
|
}
|
|
|
|
|
|
2007-08-06 13:42:52 +00:00
|
|
|
char *plugin_get_current_filename(void)
|
|
|
|
|
{
|
|
|
|
|
return current_plugin;
|
|
|
|
|
}
|