changelog

eCos操作系统源码

2004-01-09  Jani Monoses <[e-mail]>

	* include/eth_drv.h:
	* src/lwip/eth_drv.c: Update to support more than one device and
	be in sync with current lwip code.
	
2004-01-05  Gary Thomas  <[e-mail]>

	* src/net/eth_drv.c: Names of FLASH config types changed.

2003-10-09  Gary Thomas  <[e-mail]>

	* src/stand_alone/eth_drv.c: Obviate __local_enet_addr, since this
	same information is already kept in the driver structure *__local_enet_sc

2003-08-19  Gary Thomas  <[e-mail]>

	* src/stand_alone/eth_drv.c: 
	* src/net/eth_drv.c: Improve network speeds by reducing some
	overhead and also the delay when sending a packet.

2003-07-23  Nick Garnett  <[e-mail]>

	* include/eth_drv.h:
	* src/net/eth_drv.c:
	Added an ifndef to fix warnings about NBPFILTER not being
	defined. 

2003-03-20  Mark Salter  <[e-mail]>

	* cdl/eth_drivers.cdl (CYGHWR_NET_DRIVERS): Moved here from net/common.

2003-03-14  Nick Garnett  <[e-mail]>

	* src/net/eth_drv.c: Added calls to cyg_net_show_mbufs() whenever
	we see an out-of-mbufs problem.

2003-03-11  Mark Salter  <[e-mail]>

	* src/stand_alone/eth_drv.c (eth_drv_stop): Uncomment, make non-static,
	and use __local_enet_sc.
	* include/eth_drv.h: Add extern decl for eth_drv_stop.

2003-02-24  Jonathan Larmour  <[e-mail]>

	* cdl/eth_drivers.cdl: Add doc link.

2003-02-07  Jonathan Larmour  <[e-mail]>

	* src/net/eth_drv.c: If driver can't set multi-cast, indicate it's
	only a warning.

2002-08-14  Gary Thomas  <[e-mail]>

	* cdl/eth_drivers.cdl: CYGNUM_IO_ETH_DRIVERS_SG_LIST_SIZE 
	needs to be global (used by all drivers, not just NET).

2002-08-13  Gary Thomas  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_send): Print a better message
	if the scatter-gather list overflows.

	* include/eth_drv.h (MAX_ETH_DRV_SG): 
	* cdl/eth_drivers.cdl: Add control over size of scatter-gather
	data lists used to pass requests to physical layer.  Previous
	value was sometimes too small.

2002-07-31  Gary Thomas  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_start): Fix compile error when
	built with old stack (no multicast) - introduced below.

2002-07-26  Gary Thomas  <[e-mail]>
2002-07-26  Ken Cox <[e-mail]>

	* src/net/eth_drv.c (eth_drv_start): Force multicast address
	setup any time chip is reset/reconfigured.

2002-05-30  Jonathan Larmour  <[e-mail]>

	* cdl/eth_drivers.cdl: Provide CYGINT_IO_ETH_INT_SUPPORT_REQUIRED
	interface to indicate if interrupt support is required.

2002-05-28  Jonathan Larmour  <[e-mail]>

	* src/lwip/lw.diff: Remove. Obsolete.

2002-05-13  Jesper Skov  <[e-mail]>

	* cdl/eth_drivers.cdl: Moved the package's header files to
	cyg/io/eth to clean up the include root directory.

2002-04-30  Jonathan Larmour  <[e-mail]>

	* src/lwip/eth_drv.c: Update from Jani Monoses.

	* cdl/eth_drivers.cdl: Move CYGINT_ISO_STRING_STRFUNCS requirement
	into CYGPKG_IO_ETH_DRIVERS_NET where it belongs.
	Move CYGINT_ISO_STRING_MEMFUNCS requirement into
	CYGPKG_IO_ETH_DRIVERS_STAND_ALONE where it belongs.

2002-04-10   Gary Thomas <[e-mail]>

	* src/net/eth_drv.c: Only pass mutlicast addresses down to drivers.	

2002-04-05   Jani Monoses  <[e-mail]>

	* src/lwip/eth_drv.c: New file. Add eth driver support for lwip.
	* src/lwip/lw.diff: New file. diff against lwip sources.
	* src/lwip/README: New file.
	* cdl/eth_drivers.cdl: Add lwip driver support.

2002-03-09  Gary Thomas  <[e-mail]>

	* src/net/eth_drv.c: Select 'multicast_all' if too many addresses
	are required.  Also, ignore any non-multicast addresses.

2002-02-22  Hugo Tyson  <[e-mail]>

	* doc/ethdrv.sgml: New file.  SGML-ized the existing driver-doc
	file and brought the information up to date with deliver functions
	and all that.

2002-02-18  Gary Thomas  <[e-mail]>

	* cdl/eth_drivers.cdl: 
	Add new interface CYGINT_IO_ETH_MULTICAST which is used to
	insure that drivers support multicast addresses if IPv6 is used.

2002-02-15  Gary Thomas  <[e-mail]>

	* src/net/eth_drv.c: 
	* include/eth_drv.h: Add framework for multicast address support.

2002-01-30  Hugo Tyson  <[e-mail]>

	* cdl/eth_drivers.cdl (CYGPKG_IO_ETH_DRIVERS_WARN_FORCE_CONSOLE):
	CDL options removed; control is now from RedBoot.

	* src/net/eth_drv.c (START_CONSOLE): Changed to use RedBoot
	fconfig items for deciding what console to use for special debug
	messages such as "out of MBUFs" and the like.

	* src/stand_alone/eth_drv.c (LOCK_APPLICATION_SCHEDULER): New
	macro to do what it says, before passing packets to the app.
	Used in two places: when passing a received packet, and when
	passing a tx completion status back.

2002-01-28  Gary Thomas  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_init): Changes for supporting new
	FreeBSD based stack (minor API changes internal to that stack).

2001-12-12  Hugo Tyson  <[e-mail]>

	* src/net/eth_drv.c (START_CONSOLE,END_CONSOLE): New macros akin
	to those within RedBoot's internal net_io.c, for directing warning
	messages to a serial line if required.  All printf's wrapped in
	them for safety.

	* cdl/eth_drivers.cdl (CYGPKG_IO_ETH_DRIVERS_WARN_FORCE_CONSOLE):
	and ..._NUMBER, new options to direct application warning output
	somewhere useful (less harmful) for when you are debugging over
	the net.  Trying to print within a call which occurred from within
	RedBoot itself is such a recursive situation as to be fatal.
	Defaults are disabled, you must choose to set these up if
	you want to debug your app safely over the network.

2001-12-12  Hugo Tyson  <[e-mail]>

	* src/stand_alone/eth_drv.c (DIAG_DUMP_BUF_HDR,DIAG_DUMP_BUF_BDY):
	Separate macros for debug dumps of ethernet header and body.
	(eth_drv_write): Use them.
	(eth_drv_copy_recv): Do not corrupt the static data pointer, so
	that if we are erroneously called twice, at least what we pass is
	a packet, rather than random memory; do not escape to SEGV.  Guard
	against NULL sg_list[].buf pointers, in case caller out of MBUFs.
	(eth_drv_recv): Only dump packet if we actually have a buffer.
	Only forward the packet if actually have a buffer.

2001-12-04  Richard Sandiford  <[e-mail]>

	* src/stand_alone/eth_drv.c (eth_drv_read): Check that the return
	buffer is big enough to hold the packet.

2001-11-28  Jonathan Larmour  <[e-mail]>

	* doc/driver_doc: Mention preferences on how the ESA should be set.

2001-10-30  Jonathan Larmour  <[e-mail]>

	* doc/driver_doc: Add description of poll, deliver and int_vector
	driver functions.

2001-10-29  Hugo Tyson  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_tickle_devices): [CASE 106613] Rather
	than calling can_send() in the device driver directly, call our
	own eth_drv_send() function which will also dequeue a waiting
	packet if we find we indeed can_send().  This helps to recover
	from queue full situations within the if_ethersubr layer above.
	This change is belt & braces with a similar call in the code which
	drops a packet if the queue is full, in the main network code in
	net/tcpip/.../src/sys/net/if_ethersubr.c (ether_output); this
	change will recover the situation shortly even if the application
	gives up trying to send because of ENOBUFS.

2001-10-18  Jonathan Larmour  <[e-mail]>

	* cdl/eth_drivers.cdl: Rename CYGSEM_IO_ETH_DRIVERS_DEBUG to
	CYGDBG_IO_ETH_DRIVERS_DEBUG and make it a common option.
	Add separate CYGDBG_IO_ETH_DRIVERS_DEBUG_VERBOSITY config.

	* include/eth_drv.h: Don't claim BSD if it isn't.
	* include/eth_drv_stats.h: Ditto.
	* include/netdev.h: Ditto.
	* src/net/eth_drv.c: Ditto.
	Also use above CDL options to conditionalize debugging (and the
	verbosity level).
	* src/stand_alone/eth_drv.c: Use above renamed CDL options.

2001-10-11  Gary Thomas  <[e-mail]>

	* src/stand_alone/eth_drv.c (eth_drv_recv): 
	Defensive programming - insure valid packet size.

2001-10-05  Jonathan Larmour  <[e-mail]>

	* src/net/eth_drv.c: Add default implementation of min().

2001-09-26  Jesper Skov  <[e-mail]>

	* cdl/eth_drivers.cdl: Require the string functions.

	* src/stand_alone/eth_drv.c: Fix warnings.

2001-09-25  Hugo Tyson  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_dsr): Assert that the "sc" given to
	eth_drv_dsr() really is a sc from the valid list.  This can save a
	few hours debug time of a typo.

2001-09-13  Hugo Tyson  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_tickle_devices): New function to call
	can_send() on all devices to allow them to unstick from a lost
	interrupt or other wedge - it's up to the device to detect this,
	of course, can_send() does not explicitly do this, it's just used
	as an opportunity harmlessly to get control into the device so it
	can examine its world.

2001-09-05  Gary Thomas  <[e-mail]>

	* src/stand_alone/eth_drv.c (eth_drv_init): Don't initialize
	structures if ESA is zero.  This is used by low level drivers
	to indicate that a device is present, but can't currently be
	used.  The eCos driver handles this case as well.

2001-08-31  Hugo Tyson  <[e-mail]>
2001-08-31  Andrew Lunn <[e-mail]>

	* src/net/eth_drv.c (eth_drv_tx_done): Guard against a NULL key
	return - which can be caused by race conditions in the driver,
	this is the neatest fixup.  It's good defensive programming
	anyway and ASCOM's tests indicate a benefit for CASE 106059.

2001-08-22  Gary Thomas  <[e-mail]>

	* src/stand_alone/eth_drv.c: 
	printf() is no longer a part of RedBoot.  Thus all programs
	must use diag_printf() and related functions instead.

	* cdl/eth_drivers.cdl: Only enable warning/debug modes if RedBoot
	is present as the debug environment (since there are implicit
	dependencies on RedBoot functions).

2001-08-20  Jonathan Larmour  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_send): Move endif location to fix build
	error.

2001-08-17  Jonathan Larmour  <[e-mail]>

	* cdl/eth_drivers.cdl (CYGSEM_IO_ETH_DRIVERS_DEBUG): Now booldata.
	* src/stand_along/eth_drv.c (eth_drv_tx_done): Use different
	value of CYGSEM_IO_ETH_DRIVERS_DEBUG for verbosity.

2001-08-14  Gary Thomas  <[e-mail]>

	* src/stand_alone/eth_drv.c (eth_drv_write): 
	(eth_drv_tx_done): 
	(eth_drv_read): Better handling of stacking (layering) of drivers.
	RedBoot (stand alone code) is designed to call into the eCos 
	stack and these changes make sure that this is done properly
	nested/stacked.  These changes also affect the behaviour positively
	for CR 902745-CR.

	* src/net/eth_drv.c (eth_drv_send): Add locking of driver while
	actual hardware routines are involved.  Since the same driver
	can be shared by both eCos and RedBoot, it is imperative that
	additional locking (in the form of locking the scheduler) be
	employed during this window to make sure that the hardware is
	handled in complete, consistent steps.  This helps with known
	bug CR 902745-CR.

2001-08-09  Hugo Tyson  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_recv): Add a pair of
	CYGARC_HAL_SAVE_GP()/CYGARC_HAL_RESTORE_GP() on entry and exit;
	this function is intercalled between RedBoot and the application.
	A minor re-org at the ending to accommodate these macros' nesting.
	(eth_drv_tx_done): CYGARC_HAL_SAVE_GP()/CYGARC_HAL_RESTORE_GP().

	* src/stand_alone/eth_drv.c (eth_drv_tx_done): Add a pair of
	CYGARC_HAL_SAVE_GP()/CYGARC_HAL_RESTORE_GP() on entry and exit;
	this function is intercalled between RedBoot and the application.
	Also make printing message about "tx_done for other key" also
	conditional on net_debug variable - it's commonplace.
	(eth_drv_recv): CYGARC_HAL_SAVE_GP()/CYGARC_HAL_RESTORE_GP().
	(eth_drv_copy_recv): CYGARC_HAL_SAVE_GP()/CYGARC_HAL_RESTORE_GP().

2001-07-03  Gary Thomas  <[e-mail]>

	* src/net/eth_drv.c (eth_drv_recv): Better cleanup when running
	out of mbufs - could have lost mbufs before.

2001-06-21  David Woodhouse <[e-mail]>