Linux 2.4.9-ac6

Linux 2.4.9-ac6

[Alan Cox]

	ftp://ftp.kernel.org/pub/linux/kernel/people/alan/linux-2.4/

		 Intermediate diffs are available from
			http://www.bzimage.org

2.4.9-ac6
o	Update compiler requirements doc		(me)
o	Fix module count leak (I hope) in cs46xx	(me)
o	Fix sx.c warnings				(Christoph Hellwig)
o	Fix seagate.c prototypes			(Christoph Hellwig)
o	Remove non-modular stuff from mod builds	(Christoph Hellwig)
	and fix warnings
o	Fix missing return value on xirc2ps		(Christoph Hellwig)
o	Fix atmtcp MODULE_LICENSE			(Christoph Hellwig)
o	Remove various unused code			(Christoph Hellwig)
o	Switch drivers/fc4 to use module_init		(Christoph Hellwig)
o	Config file fixes				(Christoph Hellwig)
o	Fix AX.25 digipeat crash			(Thomas Osterried)
o	DECNET update					(Steven Whitehouse)
o	Fix UNUSUAL_DEV entry for eUSB SmartMedia	(Andries Brouwer)
o	Remove spare maxinefb setup			(Paul Mundt)
o	Add USB MODULE_LICENSE tags			(Greg Kroah-Hartmann)
o	Update the irq fix for the i810 audio based	(me)
	on further analysis by Doug Ledford
o	make rpm target bug-fixes			(Eli Carter)
o	Fix missing export-objs in acpi			(Keith Owens)
o	VIA ide update (support 82c576, other small	(Vojtech Pavlik)
	fixes)
o	Fix tulip bug when using MWI experimental bits	(Jeff Garzik)
o	Add MODULE_LICENSE tags to telephony		(Robert Love)
o	Add MODULE_LICENSE tags to drivers/video	(Robert Love)
o	Fix z2ram tag					(Robert Love)
o	Ask for 255 bytes of header on scsi pages	(Matt Dharm)
	| Lots of USB crap can't even get truncating right
o	Fix ver_linux for e2fsprogs 1.23		(Albert Cranford)
o	Add MODULE_LICENSE tags to zorro		(Robert Love)
o	Make __module_license static			(Keith Owens)
o	Merge some of the PPC64 submission	(Peter Bergner, Anton Blanchard, Mike Corrigan, Dave Engebretsen,
					Tom Gall, Todd Inglett, Paul Mackerras,
					Pat McCarthy, Steve Munroe, Don Reed, 
					and Al Trautman)
	| I dropped some config bits to keep stuff simpler
	| and a few files that definitely didnt follow CodingStyle
o	Merge updated gdth scsi raid driver		(Achim Leubner)
o	Remove escaped debug code from ni5010		(Frank Davies)

2.4.9-ac5
o	Make pae i386 compile again			(Russell King)
o	Add MODULE_LICENSE tagging			(me)
o	Clean up aztcd (phase 1)			(me)
o	Fix aztcd subchannel error reporting bug	(me)
o	Reformat cdu31a pending cleanups		(me)
o	Reformat cm206 pending cleanups			(me)
o	Reformat gscd pending cleanups			(me)
o	Reformat isp16 pending cleanups			(me)
o	Reformat sjcd pending cleanups			(me)
o	Reformat tpqic02 pending cleanups		(me)
o	Add tags in drivers upto and including drivers/char/*
	| lots more to add yet...
o	pl2303 oops fix					(Greg Kroah-Hartmann)
o	Sony clie updates for clie OS 4.0		(Greg Kroah-Hartmann)
o	Fix elf loader for prelink binaries		(Jakub Jelinek)
o	Make xconfig fix				(Robert Love)
o	Add reparent_to_init, fix pnp and 8139 zombies	(Andrew Morton)
o	Update Configure.help				(Steven Cole)

2.4.9-ac4
o	ns83820 driver fixes and updates		(Ben LaHaise)
o	Configure.help updates				(Steven Cole)
o	Add generic pgtable_cache_init()		(Russell King)
	| and remove pae ifdefs from init/main.c
o	Fix X.75 with new hisax drivers and an isdn	(Kai Germaschewski)
	disconnect race
o	Remove now defunct directory offset cast	(me)
o	Make several vm behaviours tunable for now	(Rik van Riel)
	| This is so we can study behaviour patterns not for
	| the long term
o	Merge an additional ide-floppy fix		(Sam Varshavchik)
	| Fixed the ide floppy I/O error funny on some drives
o	Pull dac/adc rate setting into ac97_codec.c	(me)
o	Update mips64 makefiles 			(Ralf Baechle)
o	Complete the missing bits of the proc 		(Ralf Baechle)
	infrastructure using constant HZ to userspace
	| This has been partial for a long time, with the mips tree
	| it actually needs to be completed...
o	Avoid oops in rivafb when using 15bit depth	(Steve DuChene)
	on riva128
o	Indent seagate scsi into linux format		(me)
	| Changes pending so do this in two steps..
o	pl2303 updates					(Greg Kroah-Hartmann)
o	Orinoco update					(David Gibson)
o	IRQ stack value fix				(John Byrne)
o	Enable DMA on 20268R				(Zygo Blaxell)
o	Add missing -EFAULT return to se401		(Pavel Machek)
o	Voodoo 1/2 frame buffer console			(Ghozlane Toumi)
o	Update cache size reporting errata		(Dave Jones)
o	Fix nasty oops and deadlock in i810_audio	(me)

2.4.9-ac3
o	Fix config glitch in drivers/video/Config.in	(Steven Cole)
o	Kaweth endian fixes				(Brad Hards)
o	Update the MPT fusion drivers			(Steve Ralston)
o	Possible floppy probe fix			(Paul Gortmaker)
o	Add the KT266 agp to the table			(Kris Kersey)
o	Start convering ia32 and x86_64 mtrr code	(Dave Jones)
o	Account ramdisk in out of memory code		(Russell King)
o	Possible fix for cardbus allocation failures	(Andreas Bombe)
o	Clean up other cases of const initdata		(Dave Jones)
o	Update the keyboard rate code to be more	(Dave Miller)
	flexible (needed for sparc)
o	Configure.help fixes				(Steven Cole)
o	Pegasus USB driver fixes			(Petko Manolov)
o	Fix i810 audio pops on speed changes		(Scott Herod)
o	GPIO driver for the ITE board			(Hai-Pao Fan)
o	Philips Nino port update		(Steven Hill, Pavel Machek)

2.4.9-ac2
o	Last small bits of the PPC merge		(Paul Mackerras)
o	Fix compile bugs in airport driver		(David Gibson)
o	ITE8172 ide updates				(Steve L)
o	Add i2c drivers for the ITE board		(Hai-Pao Fan)
o	AC97 register naming fix			(Ralf Baechle)
o	TI 3912 serial driver			(Harald Koerfgen, Jim Pick,
							 Steven Hill)
o	ITE general updates				(P Popov)
o	Remove double init of SGI streamable device	(Ralf Baechle)
o	Update SGI indy drivers				(Ralf Baechle)
o	Qtronix keyboard driver updates			(P Popov)
o	Add tx3192 frame buffer support			(Steven Hill)
o	MIPS frame buffer updates			(Ralf Baechle)
o	Move vino.h into driver directory		(Ralf Baechle)
o	Ocelot updates			(Jun Sun, G Lonnon, S Kranz, Steve J)
o	DDB5 updates					(Jun Sun)
o	MIPS jazz update				(Ralf Baechle)
o	SGI wd33c93 update				(Ralf Baechle)
o	Baget updates					(Ralf Baechle)
o	SNI updates					(Ralf Baechle)
o	Alchemy Au1000 support				(P Popov)
o	MIPS eval board updates		(Ralf Baechle, Carsten Langgaard)
o	Update Decstation serial support		(Maciej W. Rozycki)
o	NEC Vrc5477 audio driver			(Steve L)
o	General MIPS32 updates		(Jun Sun, Ralf Baechle, Matt Porter,
					 Kevin Kissell, Carsten Langgaard,
					 Jan-Benedict Glaw)
o	MIPS scsi updates				(Ralf Baechle)
o	Notifier signal oops fix		(Benjamin Herrenschmidt)

2.4.9-ac1
o	Merge the fat and iso changes from 2.4.9
o	Merge the sunrpc changes from 2.4.9
o	Merge (hopefully correctly) the nfs changes
o	Switch to the 2.4.9 emu10k1 driver
o	Merge vfs directory type changes
o	Merge other oddments
	- This leaves min/max and the vm/buffer changes
	  both of which are pretty dubious anyway
o	lock_kiovec page unwind fix			(Velizar B)
o	do_swap_page recheck pte before failing		(Linus, Jeremy Linton)
o	do_swap_page doesn't mkwrite when deleting	(Linus)
	| From 2.4.9 with extra comments etc		(Hugh Dickins)

2.4.8-ac12
o	Merge the majority of 2.4.9 except
	- min/max mess
	- fat/isofs changes
	- drm changes (some collisions with other
			fixes)
	- vm/buffer handling changes
	- emu10k1
	- vfs directory type changes
	- nfs/nfsd/sunrpc
	I'm trying to make sure I can keep this testable
	as 2.4.9 vanilla isnt being stable on my test sets 
	This is basically a merge of all the "boring" bits.

o	Update usb network fixes			(Herbert Xu)
o	HID id matching fix				(Pete Zaitcev)
o	ARM i/o fixes					(Russell King)
o	ARM alignment trap fixes			(Russell King)
o	ARM softirq fix					(Russell King)
o	Miscellaneous arm updates			(Russell King)
o	Configure.help updates			(Andrzej Krzysztofowicz,
							 Steven Cole)
o	Prefetchw macro fix				(Andreas Franck)
o	Kill another bogus wbinvd macro in ACPI		(Dave Jones)
o	Switch PnPBIOS to spinlocks and irq off		(me)
o	Large PPC merge					(Paul Mackerras)
o	Remove surplus macintosh rtc printks		(Paul Mackerras)
o	Add modem power features to the mac serial 	(Paul Mackerras)
o	Powerbook trackpad lockup fix			(Paul Mackerras)
o	Adb updates					(Paul Mackerras)
o	Fix serial port base offsets			(Paul Mackerras)
o	PPC sysctl code to use PPC32 not __powerpc__	(Paul Mackerras)
	| The latter is both 32 and 64 bit...
o	PowerMac pmu updates				(Paul Mackerras)
o	S/390 3270 driver update			(Richard Hitt)
o	MIPS docs update				(Ralf Baechle)
o	Update mips maintainers entry			(Ralf Baechle)
o	Update mips configure.help			(Ralf Baechle)
o	Add some mips pci ids				(Ralf Baechle)
o	DECstation turbochannel update			(Maciej W. Rozycki,
							 Ralf Baechle)
o	MIPS64 update					(Ralf Baechle)

2.4.8-ac11
o	Remove bogus wbinvd macro in ACPI		(Dave Jones)
o	Fix highmem high page races			(Ben LaHaise)
o	SCSI generic driver updates			(Doug Gilbert)
o	Remove unneeded fusion init call		(Adam J Richter)
o	Fix init/const clashes in pci			(Dave Jones)
o	Merge newer qlogic fc firmware set		(Ricky Beam)
o	Add NEC DV5800A to the blacklist for DMA	(Matt Domsch)
o	Add passive ISDN over USB support		(Frode Isaksen,
							 Kai Germaschewski)
o	Fix aha1542 flags type				(Andi Kleen)
o	Macsonic kmalloc type fixes			(Dave Jones)
o	lock clear_page_tables versus kswapd		(Ben LaHaise)
o	Honour PnPBIOS region reservations		(Gerd Knorr)
o	Made them __init and fixed problems when the	(me)
	resource declaration spanned 0x100
o	Fix a rio serial type warning			(me)
o	Add another byteswapped Vaio BIOS		(Roger Luethi)
o	USB serial tidy ups				(Greg Kroah-Hartmann)
o	Fix dev->actconfig changing in probe in USB	(Michael Stickel)
	causing later errors

2.4.8-ac10
o	Fix the USB device timeout problem		(Pete Zaitcev)
o	rio usb locking fixes				(Oliver Neukum)
o	Fix vm86 v segment reload part 1		(Andi Kleen)
	| Still not all sorted yet
o	Airo driver update				(Javier Achirica)
o	Fix bugs in usb skeleton driver			(Greg Kroah-Hartmann)
o	Add support for USB clie serial devices		(Greg Kroah-Hartmann)
o	USB config fix for serial debug			(Greg Kroah-Hartmann)
o	Add support for massworks id75 usb		(Greg Kroah-Hartmann)
o	Rationalise lvm version numbers			(A J Lewis)
o	LVM locking changes				(Joe Thornber)
o	Pull pv flush out of lvm ioctls			(Heinz Mauelshagen)
o	Switch to lv_v5_t structures/types in LVM		(Heinz Mauelshagen)
o	Defer LVM I/O when moving an extent		(Joe Thornber,
							 Andreas Dilger)
o	Fix bmap/blkszget issues in LVM			(Andreas Dilger)
o	Tidy naming/devfs registrations for LVM		(Patrick Caulfield)
o	Protect LVM snapshot flag removal, don't
	extend or reduce dropped snapshots, 
o	Handle ELF loader setup arg pages failures	(Evgeny Polyakov)
o	Add udelays to cmpci to see if it fixes		(me)
	the problems a few people have
o	Put config hooks in to make qlogicfc firmware	(me)
	optionally loadable for weird hardware
	| Needs a suitable firmware file adding ..
o	Update pci.ids for a couple of parisc things	(Helge Deller)
o	Irda warning fixes				(Pete Zaitcev)
o	Squash smp race in dsbr100 driver		(Oliver Neukum)
o	Console locking fix on VT_DISALLOCATE race	(Jani Jaakkola,
							 Andrew Morton)
o	ipv4 raw socket oops fix			(Octavian Cerna)
o	Actually use our msr register defines 		(Dave Jones)
o	Correct polish translation info typos		(Steven Cole)
o	mm compile warning fix				(Rik van Riel)
o	Updated 3ware driver				(Adam Radford)
o	Add Sharp PC-RJ/AX to bad apm list		(Arjan van de Ven)

2.4.8-ac9
o	Possible usb -110 error fix			(me)
o	Page laundering fix				(Rik van Riel)
o	Add another vaio to the byteswap list		(Martin Mueller)
o	Fix typos in ide blacklist			(Arjan van de Ven)
o	UMSDOS split directory entry handling		(Istvan Varadi)
o	Update Configure.help further			(Steve Cole)
o	Fix bogus mtrr warning on dual pentiums		(Dave Jones)
o	Update scsi tape driver				(Kai Mäkisara)
o	Change memory probe constants on AWE32/64	(Dave Fennell)
o	Tiny endian reiserfs fix (just cosmetic)	(Jeff Mahoney)
o	Make the md resync delay message informative	(Corin Hartland-Swann)
	rather than scary
o	USB printer fixes				(Oliver Neukum)
o	Add the SIS735 to the SiS AGP			(Adrian)
o	Kill bogus export_objs entry in lib/Makefile	(Keith Owens)
o	VIA rhine fixes					(David Woodhouse)
o	Next set of superblock changes			(Al Viro)
o	Don't reissue a pid that has a tgid matching	(Dave McCracken)
	it still in circulation

2.4.8-ac8
o	Fix double mount hang on scsi cdrom, i2o or lvm	(Al Viro)
o	Fix oops in msdos/umsdos			(OGAWA Hirofumi)
o	Fix qnxfs hang					(Serguei Tzukanov)
o	Add missing Alpha ksyms				(Marc Zyngier)
o	USB oops fixes					(Pete Zaitcev)
o	Apply same fix to kaweth			(me)
o	Fix off by one in pcigart			(Andreas Schwab)
o	Fix dasd leak					(Al Viro)
o	page reactivate correction			(Rik van Riel)
o	Add the 104K to the byteswapped minutes bug	(Daniel Caujolle-Bert)
	list
o	Missing USB config items			(Mike Castle)
o	Fix i2o systab send id order			(Klaus Beyer)
o	VMA merging fixups				(Ben LaHaise)
o	Update ntfs 					(Anton Altaparmakov)
o	Speed up ext2 readdir/stat			(Ted Tso)
o	Comment/docbook fixups				(Dave Jones)
o	Add missing netif_wake_queue calls to USB	(Herbert Xu)
	network drivers
o	Fix cramfs to use kmap				(Herbert Xu)
o	Fix NFS client atomic_dec_and_lock symbol	(Trond Myklebust)
o	Make wake_up_interrutible_sync usable in 	(Jeremy Elson)
	modules
o	Update PPC for kbd_rate support			(Paul Mackerras)
o	Next block of Configure.help tidying	(Andrzej Krzysztofowicz)
o	Next block of superblock cleanup		(Al Viro)
o	Update hp scanner driver for USB		(Oliver Neukum)
o	Clean up ibm partition code			(Al Viro)

2.4.8-ac7
o	Further small DRI/AGP updates			(Jeff Hartmann)
o	Update master makefile to force offset.h/	(Andi Kleen)
	version.h/depend order (needed for x86-64)
o	Merge x86-64 architecture port code		(Andi Kleen and co)
o	Merge ixj update				(Craig Southern)
o	Further ixj cleanup/merge tweaks		(me)
	| Ie don't blame him ;)
o	Grand config file cleanup		     (Andrzej Krzysztofowicz,
							Steven Cole)
o	Add another byteswap vaio			(Ray Lee)
o	Correct partition check oops fix		(Kevin Flemming)
o	via82cxx IDE DMA updates		 	(Vojtech Pavlik)
	| Enable 8231/8233 support, handle slightly
	| out of spec PCI from 1.3GHz/12.5x Athlon
o	Rip min/max use entirely out of isdn		(Kai Germaschewski)
	| To handle 2.4.9 compat disaster
o	Update K6 bug url				(André Dahlqvist)
o	Possible fix for trix ad1848 fail		(me)
o	Add pci quirk warning for AMD766 errata 22	(me)
	| Based on multiple "yes noapic fixed my
	| dual athlon" reports.

2.4.8-ac6
o	Pull Linus buffer.c/mem fixes into 2.4.8-ac	(Rik van Riel)
o	Make writeout smoother on zone specific		(Marcelo Tosatti)
	shortages
o	Fix an md oops					(Neil Brown)
o	AD1848 isapnp handling				(Miguel Freitas)
o	sr_ioctl capacity reporting fix			(Jens Axboe)
o	Starfire update					(Ion Badulescu)
o	cmpci update					(Carlos Gorges)
o	UDF update, fix delete BUG() trap		(Ben Fennema)
o	Move nmi defines so asm/irq.h isnt required	(Russell King)
o	Add experimental requirement to intermezzo   (Andrzej Krzysztofowicz)
o	Fix missing include 				(Tom Rini)
o	Small eepro100 test updates			(Arjan van de Ven)
o	Reiserfs transaction tracking update		(Chris Mason)
	| Speeds up O_SYNC and fsync
o	Update vfs_permission, handle root exec		(Christoph Hellwig)
	weirdness
o	First piece of the CyberPro 5050 audio merge	(Peter Wächtler)
o	Tidy i810_audio cornercases of OSS compliance	(Laurent Pinchart)
o	IDE cdrom blacklist updates for DMA		(Matt Domsch)
o	Merge some i810 updates 			(Doug Ledford)
o	Fix bug in i810 device removal 			(me)
	| fortunately nobody yet has multiple ICH audio in one box 8)
o	Natsemi gige driver				(Ben LaHaise)
o	Allow more I/O addresses on msnd_pinnacle	(Steve Sycamore)
o	Correct sys_tz definition			(Andi Kleen)
o	Remove unused prototypes from md		(Andi Kleen)
o	Fix flags wrong types in usb			(Andi Kleen)
o	Fix assorted wrong flags, add x86_64 defines	(Andi Kleen)
o	Fix flags types in i2o				(Jes Sorensen)
o	Add another vaio byteswap case			(Stelian Pop)
o	Add devfs support to usb scanner		(Yves Duret)
o	Fix an ess solo warning				(Christoph Hellwig)
o	Fix flags types in 3ware driver			(Jes Sorensen)
o	Fix generic serial warnings			(Christoph Hellwig)
o	Fix flag types in firewire drivers		(Jes Sorensen)
o	Fix flag types in dz serial			(Jes Sorensen)
o	Extend short name handling in fat based fs's	(OGAWA Hirofumi)
o	Fix flag types in n_r3964			(Jes Sorensen)
o	x86_64 ifdef hooks for raid and fbcon		(Andi Kleen)

2.4.8-ac5
o	Next batch of IDE driver updates		(Andre Hedrick)
	| qd6580 driver becomes a qd65xx driver
	| 80pin cable detect for serverworks on Dell
	| Mode5 on SIS chipsets
o	Handle ARM mmap were FIRST_USER_PGD_NR is not	(Russell King)
	zero
o	Make the sl82c105.c code common between ARM	(Paul Mackerras)
	and PPC
o	Update cisco hdlc handling in the isdn layer	(Bjoern Zeeb)
o	Add called party number to isdn tty emulation	(Jan Oberlaender,
							 Kai Germaschewski)
o	Eicon warning fix				(Kai Germaschewski)
o	Tiny agp cleanup in severworks code		(Mike Harris)
o	Switch to 2.4.8 nr_free_buffer_pages		(Rik van Riel)
o	Change bootmem bitmap setup			(Rik van Riel)
o	Unlazy the queue movement when we touch
	inactive cache pages (VM balance assumed this)	(Rik van Riel)
o	Update the orinoco drivers			(David Gibson)
o	Update natsemi driver (experimentally anyway)	(Tim Hockin)
o	Update hpt366 blacklists			(Kevin Fleming)
o	Reclaim buffer cache into inactive list when	(Rik van Riel)
	it is too large
o	Documentation tidy ups				(Steven Cole)
o	Switch map_user_kiobuf to use down_read		(Ben LaHaise)

2.4.8-ac4
o	ADFS date/time computation fix			(Russell King)
o	Add ALS120 ident to ns558 joystick		(Filip Van Raemdonck)
o	Make Reiserfs endian and alignment safe		(Jeff Mahoney)
	| Fixes IA64 indirect alignment, S390 alignment
	| Big endian
	| Update inode generator
o	Enable input drivers on ARM			(Russell K
o	Add intermezzo file system kernel side		(Peter Braam and co)
o	First blocks of ppc64 merge	(Paul Mackerras, Anton Blanchard,
					 Tom Gall and the IBM PPC 64 team)
o	Fix return value bug in mac nvram driver	
o	Make oom killer kill all threads of a set	(Eric Lammerts)


2.4.8-ac3
o	Update ISDN cvs idents				(Kai Germaschewski)
o	Update isdn experimental idents			(Kai Germaschewski)
o	Add Compaq Phoenix 4.06 to irqsafe APM		(Arjan van de Ven)
o	Remove rockridge printk generating too much	(Mikael Pettersson)
	debug output
o	Add a needed include of delay.h			(Arjan van de Ven)
o	Revert access DAC change 			(me)
	| Breaks in some cases
o	ARM documentation, credits, maintainer updates	(Russell King)
o	Atyfb build fixups				(Stelian Pop)
o	Configure.help merges				(Steven Cole)
o	Initial merge of dpt_i2o.c		(Deanna Bonds, Bob Pasteur, 
						 Karen White, Mark Salyzyn)
o	Teach i2o_core to skip dpt cards		(me)
o	Sony Pi driver update				(Stelian Pop)

2.4.8-ac2
o	Fix suspend/resume bugs in eepro100		(Arjan van de Ven)
o	Fix missing spec required delays in PCI PM	(me)
o	Disable PM on eepro100, it breaks even with	(me)
	those fixed
o	Resynchronize Configure.help			(Steven Cole)
o	Further superblock handling updates		(Al Viro)
o	Fix various GPL misreferences		(Andrzej Krzysztofowicz)
o	Updated A20 gate switching code to handle	(Peter Anvin)
	odd (Olivetti etc) and no legacy boxes
o	Fix expand_stack race				(Manfred Spraul)
o	Fix a malloc failure path in ipc		(Manfred Spraul)
o	Revert 2.4.8 scsi_lib change (hangs ide-scsi	(Arjan van de Ven)
	on some drives)
o	Fix binfmt elf strlen->strnlen_user		(me)
o	Add additional checking to binfmt_elf		(me)
	| Last two based on Solar Designers 2.2 work
o	Make pnp_bios dock thread exit when it finds 	(me)
	docking isnt supported
o	Ext3 file system updates			(Andrew Morton,
							 Stephen Tweedie)
o	Merge common components of uhci drivers		(Brad Hards)
o	Remove crud from media Makefile			(Keith Owens)
o	Fix build on Alpha				(David Gilbert)
o	DRM warning/oops fix				(Arjan van de Ven)
o	Fix i810 audio return funny			(Laurent Pinchart)
o	Fix ldm partition checking			(Richard Russon)
o	Agpgart typo fix				(Mike Harris)
o	Revert emu10k changes in 2.4.8, wait until the	(me)
	maintainers actually have debugged code and
	want an update
o	Fix a compile fail case for the 53c700 driver (Andrzej Krzysztofowicz)
o	ARM core and video updates			(Russell King)

2.4.8-ac1
o	Merge Linus 2.4.8
	- Skip VM changes for now
o	Fix sblive build problems			(Rui Sousa)
o	Add Fernando Fuganti to credits			(Fernando Fuganti)
O	Revert printk return change			(Andrew Morton)
o	Add drm-4.0 to mod_subdirs			(Brian Dushaw)
o	Bluez bluetooth updates				(Maksim Krasnyanskiy)
o	Fix serverworks AGP memory leak			(Hugh Dickins)
o	Update DRM 4.1 for Alpha AGPGart support	(Jay Estabrook,
							 Jeff Hartmann)
o	Fix depca crash on unload			(Peter Denison)

2.4.7-ac11
o	Fix dumb bug in the bootflag handling code	(Randy Dunlap)
o	Compaq FC update (makefile clean too)		(Charles White)
o	Add Matt Domsch to the credits			(Matt Domsch)
o	Update Randy Dunlap's contact info		(Randy Dunlap)
o	SA1100 updates					(Russell King)
o	Add PnP support to sf16i			(Ladislav Michl)
o	Acorn drivers update				(Russell King)
o	Fix clashing 'cams' symbol			(Keith Owens)
o	Fix AGP memory leak on serverworks		(Jeff Hartmann)
o	Fix an off by one in shmem.c			(Hugh Dickins)
o	Revert incorrect ATM change			(Mike Westall)
o	DRM makefile fixes				(Keith Owens)
o	Next batch of superblock work			(Al Viro)
o	Fix in2k oopses					(Francois Romieu)
o	Add reboot by SMP reset				(Matt Domsch)
o	Add pm support for laptops that resume the	(Matt Domsch)
	mouse port even if was disabled before..
o	Add DMI quirks for all modern dell systems	(Matt Domsch)
o	Inspiron 4000 needs irqs on for APM		(Arjan van de Ven)
o	YMFPCI fixes, big endian fixes			(Pete Zaitcev)
o	Add winbond W83971D ac_97 codec to tables	(Andrey Panin)
o	pcnet32 oops fix and leak fix			(Paul Gortmaker)
o	Use static inline in sound drivers		(me)
o	Fix DRM build/makefile bugs from merge	(Andrzej Krzysztofowicz)
o	Avoid lp module being pinned down if console	(Tim Waugh)
	enabled

2.4.7-ac10
o	Fix up USB merge mess				(Pete Zaitcev)
	| Fixes a possible USB deadlock
o	IRDA update					(Dag Brattli)
o	Merge DRM for XFree 4.1.x			(XFree86 and others)
o	Update freevxfs idents				(Christoph Hellwig)
o	Fix a bug in access() checks on X_OK with	(Christoph Hellwig)
	DAC ovveride
o	Add another intel bios ident with bad $PIR	(Arjan van de Ven)

2.4.7-ac9
o	Print warnings about buggy 440GX $PIR tables	(Arjan van de Ven)
o	Update ARM softirq code				(Russell King)
o	Compaq FC controller update			(Charles White)
o	Update ARM integrator platform			(Russell King)
o	Miscellaneous ARM fixes				(Russell King)
o	ARM io function updates				(Russell King)
o	Remove duplicate Configure.help items		(Steven Cole)
o	Update ARM shark platform			(Russell King)
o	ARM anakin platform				(Russell King)
o	Allow swap < 2*ram				(Rik van Riel)
o	Set page format bit for scsi-2 tape		(Kai Makisara)
o	Fix compile with shmfs disabled			(Christoph Rohland)
o	RME Hamerfall audio driver			(Guenter Geiger)
o	Further UML fixes				(Jeff Dike)
o	Syncppp fix					(Bob Dunlop)
o	Farsync update					(Bob Dunlop)
o	UML network driver update			(Jeff Dike)
o	Revert aic7xxx makefile changes			(Keith Owens)
o	DaveJ has received enough Rise cpu reports	(Dave Jones)
o	Clean up building without procfs	    (Andrzej Krzysztofowicz)
o	Riscom compile fix			    (Andrzej Krzysztofowicz)

2.4.7-ac8
o	Kill accidental bit of S/390 merge I meant	(Bill Nottingham)
	to skip (hotplug should be working again now)
o	Fix host_info_lock namespace on ieee1394	(Keith Owens)
o	Fix duplicate rio serial init			(Keith Owens)
o	Fix dead init_zoran_cards symbol		(Keith Owens)
o	Don't define EXPORT_SYMTAB in cmpci		(Keith Owens)
o	Don't define EXPORT_SYMTABL in sisfb		(Keith Owens)
o	Fix warnings in ess_solo1 			(me)
o	Fix deadlock in moxa mxser driver		(Christophe Barbé)
o	First of many needed devfs race fixes		(Al Viro)
o	Windows 2000 vfat name mapping fix		(Wolfram Pienkoss)
o	Clean up ubd, CONFIG_IOMEM->CONFIG_MAPPER	(Jeff Dike)
o	Fix a UML crash, make uml devices pluggable	(Jeff Dike)
o	Use page cache in hostfs, fix UML stat64 bits	(Jeff Dike)
o	Complete UML ppc support merge			(Chris Emerson)
o	Complete UML Configure.help			(Bill Stearns)
o	Add rep nop so the poor old Pentium IV doesnt
	go thermal slowdown every long mdelay		(Arjan van de Ven)
o	Fix various invalid Config script items		(Christoph Hellwig)
o	SYS5fs BSD style symlink support (SCO etc)	(Christoph Hellwig)
o	Fix PnPBIOS reporting on io v mem		(Andrey Panin)
o	Fix atyfb compilation problems with vaio bits	(Keith Owens)
o	Make HP support in AMI Megaraid run time 	(Michael Johnson)
o	Fix an ext3 buffer credit accounting bug	(Andrew Morton)
o	Add hardware volume control support to ALi	(Matt Wu)
o	Clean up the above a little for non ALi, fix	(me)
	rmmod crash
o	Adaptec scsi update (6.2.1)			(Justin Gibbs)
	| + gcc 3.0 fixes
o	Handle broken PIV SMP tables			(Maciej Rozycki)
o	Switch to static inline on ARM subtree		(Russell King)
o	SHMfs updates, race fix				(Christoph Rohland)

2.4.7-ac7
o	Import safer bits of 2.4.8pre
	- skipped vm hackery, buffer hacks
	- skipped S/390 (my tree should be newer anyway)
o	Fix duplicates in ldm.h				(Matthew Gardiner)
o	ARM updates					(Russell King)
o	Remove qlogicfc_asm (license clash question + we(Jes Sorensen, me)
	| are shipping qlogic release candidate not final
	| firmware.  Set card to use firmware in flash
o	Example iomem driver + fixes for UML		(Greg Lonnon)
o	Clean up unique UML machine id code		(Henrik Nordstrom)
o	Small UML bug fixes				(Jeff Dike)
o	Fix UML uaccess macro bug			(Jeff Dike)
o	Don't refuse unneccessarily on remount with	(Petr Vandrovec)
	data= option with ext3
o	tdfx driver cleanups				(Paul Mundt)
o	Replace masq crash fixes with corrected and	(Rusty Russell)
	"official" versions.
o	Fix umsdos symlink bug				(Delbert Matlock)
	
2.4.7-ac6
o	Switch JFFS to use completions			(me)
o	Fix wrong unlock in try_to_sync_unused_inodes	(Petr Vandrovec)
o	Update UML to compile again			(Jeff Dike)
o	Clean up all drivers with clashing names for	(me)
	complete, wait_for_completion
o	Some gcc 3.0 warning cleanups
o	Add SCO AFS awareness (read only) to sysv	(Christoph Hellwig)
o	Netfilter crash fixes				(Mark Boucher)
o	Second batch of superblock race/cleanup work	(Al Viro)

2.4.7-ac5
o	Resolve ext3 and superblock change incompat	(me, Al Viro)
	| Fixes hang on journal recovery
o	Fix freevxfs leak				(Andries Brouwer)
o	Add another eepro100 ident			(Matt Wilson)
o	Further tweakes to make rpm based on 
	suggestions by Keith Owens			(Keith Owens, me)
o	Fix ISA dma range check bug in cs89x0		(Paul)
o	Fix ISA dma range check on 3c505		(Paul)
o	Switch md driver to use completions		(Neil Brown)
o	Add more sanity checks to the dmi scanner	(me)
o	Fix hash sign assumptions in reiserfs		(Jeff Mahoney)
	| IMPORTANT: this makes things consistent, it also means
	| that if you have chars > 127 in file names _and_ you are
	| running unsigned char little endian default (probably a tiny tiny
	| number of mips users only) you will need to archive and recreate
	| those files. X86 users are _NOT_ affected.
o	Add the infrastructure for ac97_ops ready for	(me)
	digital audio etc

2.4.7-ac4
o	Fix inode cache shrinkage problems in 2.4.7	(Al Viro)
	| This should cure the problem where it gets 
	| really slow over time. Its not the final fix
o	Make aironet compile again			(Arjan van de Ven)
o	Fix an incredibly stupid i2o_scsi bug causing	(me)
	crashes with the adaptec 2100 and other stuff
o	Fix memory corruption if using gcc 3 and serial	(Thomas Hood)
	probing fails
o	Fix out of memory handling with raid		(Neil Brown)
o	Fix a raid mishandling bug on errors		(Neil Brown)
o	Add promise 20268 software raid card idents	(me)
o	Fix leaktek winview601 problems			(Leandro Lucarella)
o	Merge ntfs 1.1.16				(Anton Altaparmakov)
o	Avoid panic when reiserfs attempts to mount	(Nikita Danilov)
	invalid superblock
o	Error rather than panic on journal replay I/O	(Chris Mason)
	error in reiserfs
o	Update atp870u driver				(Wittman Lee)
o	First batch of superblock handling cleanup	(Al Viro)
o	Restore module oops dumping			(Kai Germaschewski)
o	NFSD update					(Neil Brown)
o	Fix ieee1394 sleep with spinlock held		(Andi Kleen)
o	Add AMD 760-MP to the Agp table			(me)
o	Rip out zillions of duplicated -ESPIPE		(Christoph Hellwig)
	llseek methods for a common one
o	Fix qlogic direction flag handling in 2.4	(Jeff Andre)
o	Clear inode->i-blocks on deletion in reisefs	(Nikita Danilov)
o	Merge rest of S/390 tty driver fixes		(Ulrich Weigand)
o	Finish adapting S/390 to new softirq code	(Ulrich Weigand)
o	Remove accidental duplicate block in Makefile	(Ulrich Weigand)
o	Update DASD drivers				(Ulrich Weigand)
o	Further pnpbios fixes				(Andrey Panin)
o	Switch pnp to use slab not malloc.h	(Arnaldo Carvalho de Melo)
o	Switch parport_cs to slab not malloc.h	(Arnaldo Carvalho de Melo)
o	Further config cleanups				(Steven Cole)
o	Fix make spec with no .version			(Keith Owens)
o	Further ipchains fixes				(Rusty Russell)
o	Add procfs info option for reiserfs stats	(Nikita Danilov)
o	Remove dead code from the reiserfs tree		(Jeff Mahoney)
o	Quota updates					(Jan Kara)
o	FreeVxFS leak fixes, allow block sizes != 1024	(Christoph Hellwig)
o	Improve serial_cs reporting			(Jonathan Corbet)
o	Add v7 fs sanity checks to the sys5 fs code	(Linus Torvalds,
				Christoph Hellwig, Al Viro, Andries Brouwer)

2.4.7-ac3
o	Add "make rpm" target				(me)
	| Remember to install not upgrade kernel rpms
	| as you will want to keep old ones around too
o	Fix FPU emulation breakage			(Brian Gerst)
o	Fix minix subpartition handling			(Andries Brouwer)
o	Add another odd vaio bios to the apm list	(Robert Dunlop)
o	Further Configure.help updates			(Steven Cole)
o	Update usb configure and help texts		(Brad Hards)
o	Fix kmem read loop bug				(Hugh Dickins)
o	Two warning fixes				(Art Haas)
o	Fix missing icmp errors for udp			(Alexey Kuznetsov)
o	Fix 3c59x module load problem			(Hugh Dickins)
o	Update pwc driver				(Nemosoft Unv)
o	Remove old non kernel code from reiserfs tree	(Nikita Danilov)
o	Remove unneeded reiserfsck code from tree	(Nikita Danilov)
o	Replace checks in CONFIG_REISERFS_CHECK with	(Nikita Danilov)
	cleaner macros
o	Add const's correct formatting/typechecking	(Nikita Danilov)
o	Clean up do_reiserfs_warning macro		(Jeff Mahoney)
o	Add Randolph Chung to CREDITS			(Matthew Wilcox)
o	Merge some pa risc tree Configure.help		(Matthew Wilcox)
o	Update the pa risc tree docs 			(Matthew Wilcox)
o	Add pa-risc keyboard drivers	(Debacker Xavier, Marteau Thomas,
					 Djoudi Malek, Philipp Rumpf, 
					 Alex deVries)
o	Update lasi ethernet drivers for pa-risc
o	Correct acenic check for parisc to hppa
o	HPPA port requires pci-setup
o	Add headers for som binary loader		(Matthew Wilcox)

2.4.7-ac2
o	Update motion eye driver, fix stack and hang	(Hugh Dickins)
	problems
o	Fix sigsuspend bug on Alpha			(Richard Henderson)
o	CRIS port updates				(Bjorn Wesen)
o	Add support for Win2K dynamic disk partitions	(Richard Russon,
							 Anton Altaparmakov)
o	ISDN multilink fix				(Karsten Keil)
o	HFC PCI support is now ITU approved		(Werner Cornelius)
o	Update sonypi driver				(Stelian Pop)
o	Fix sleep with irq disabled bug in act2000 	(Kai Germaschewski)
o	Farsync driver cleanups				(Bob Dunlop)
o	Remove surplus tty==NULL checks in isdn tty	(Rob Radez,
							 Kai Germaschewski)
o	Remove sleep with irq disabled in icn driver	(Kai Germaschewski)
o	Remove sleep with irq disabled bug in isdnloop	(Kai Germaschewski)
o	Remove bcopy and duplicate snprintf from eicon	(me)
	driver
o	Remove bzero from eicon driver, use skb_purge	(Armin Schindler)
o	Updated NinjaScsi driver			(YOKOTA Hiroshi)
o	Fix a jffs2 memory leak				(David Woodhouse)
o	Updated bttv driver. Corruption on close fixes	(Gerd Knorr)
o	Audio driver for bt878 tv chip onboard audio	(Gerd Knorr)
o	Correct nvram byte range limit			(Aaron Rendahl)
o	Include linux/smp_lock in shmem			(Clemens)
o	Add DMI/apm support for broken bioses reporting (Marc Boucher)
	battery life wrong endian
o	Fix dumb bug in the sbf code			(me)
o	Fix pnp shutdown oops				(me)

2.4.7-ac1
o	Alpha fixes and updates				(Jay Estabrook)
	- ioremap fixes for AGP on Alpha
	- pci_iommu support for AGP on Alpha
	- AGP arch support code for Alpha UP1x00
	- Fix instruction fault bug with old Jensen/UDB
	  PALcode
o	Further alpha pci iommu cleanups/fixes		(Ivan Kokshaysky)
o	Fix DAC960 for completion change		(Jens Axboe)
o	Fix data corruption on ide tape while reading	(Pete Zaitcev)
	near EOF
o	Switch up_and_exit to complete_and_exit		(David Woodhouse)
o	Fix crash on scsi request alloc failure in sd	(Rasmus Andersen)
o	Handle scsi register failure in ultrastor	(Rasmus Andersen)
o	Clean up ioremap as u32 stuff in ibmtr		(me)
o	Fix max_sector cleanup in paride		(Andrea Arcangeli)
o	
o	Bring IDE floppy up to date with maintainer	(Gunther Mayer)
o	Remove escaped junk from Makefiles		(Christoph Hellwig)
o	Fix cs46xx checks on ioctl calls		(Simon Horman)
o	Update cpqfc driver				(Charles White)
o	NTFS fixes			(Anton Altaparmakov, Rasmus Andersen)
o	Further FATfs updates				(OGAWA Hirofumi)
o	S/390 network driver updates			(Ulrich Weigand + )
o	Core S/390 changes to get it building again	(Martin Schwidefsky)
o	Fix S/390 tree asm blocks to compile with their	(       and
	newest gcc set
o	Update S/390 documentation			(other IBM folks)
o	Update S/390 tape driver			(       "	)
o	Update S/390 console driver			(	"	)
o	Don't do net hotplug during booting		(	"	)
o	Remove tools stuff from S/390 tree		(	"	)
o	Update S/390 irq/softirq handling		(	"	)
o	Add shared kerne support for S/390 VM		(	"	)
o	Update cpqarray driver				(Charles White)
o	Add missing barrier() calls in serial drivers	(me)
o	Fix menuconfig return code on small screen	(Herbert Xu)
o	Small UML fixups				(Jeff Dike)
o	Add COW support to UML block driver		(Jeff Dike)
o	USB network driver updates/fixes		(David Brownell)
o	PnP parsing bug fix				(Andrey Panin)
o	Fix UFS checking of NULL error cases		(Andreas Dilger)
o	Further shmem bits				(Christoph Rohland)
o	Update the mmap changes to handle OSF emulation	(Maciej Rozycki)
o	Fix failed register handling in g_NCR5380	(Rasmus Andersen)
o	Correct errors in devices.txt examples		(Andreas Dilger)
o	Synclink driver update				(Paul Fulghum)
o	Fix scc region requests for latches		(Rob Turk)
o	Fix FB_ACTIVATE_NOW/VBL handling on aty128fb	(Andreas Hundt)
o	Correct CRTC_OFFSET_CNTL on aty128fb		(Andreas Hundt)
o	Add Devfs support to rio500			(Gregory Norris
							 Greg Kroah-Hartmann)
o	3c59x driver updates				(Andrew Morton,
							 Donald Becker)
o	Fix ircomm handling with some mobile phones	(Andrea Arcangeli,
							 Dag Brattli)
o	Clean up resource handling in ibmtr driver	(Rasmus Andersen)
o	Handle out of memory for device alloc in gdth	(Rasmus Andersen)
o	Fix mixer leak in sb driver			(Mike Galbraith)
o	Make printk use vsnprintf			(Andrew Morton)
o	Fix bug in atm_do_connect_dev (bogus EINVALs)	(Germán González)
o	Add Sony DSC-575 to unusual devices usb list	(Denis Benoit)
o	Add gemtek pci radio card driver		(Vladimir Shebordaev)

 
---
Alan Cox <alan@lxorguk.ukuu.org.uk>
Red Hat Kernel Hacker
& Linux 2.2 Maintainer                        Brainbench MVP for TCP/IP
http://www.linux.org.uk/diary                 http://www.brainbench.com

Re: Linux 2.4.9-ac6

[Keith Owens]

On Mon, 3 Sep 2001 02:50:47 +0100, 
Alan Cox <laughing@shared-source.org> wrote:
>2.4.9-ac6

-ac4 added to arch/i386/kernel/setup.c::display_cacheinfo()

        if (c->x86_vendor == X86_VENDOR_CENTAUR && (c->x86 == 6) &&
                (c->x86_model == 7) || (c->x86_model == 8)) {

That should probably be

        if (c->x86_vendor == X86_VENDOR_CENTAUR && (c->x86 == 6) &&
                ((c->x86_model == 7) || (c->x86_model == 8))) {

Re: Linux 2.4.9-ac6

[Dave Jones]

On Mon, 3 Sep 2001, Keith Owens wrote:

> -ac4 added to arch/i386/kernel/setup.c::display_cacheinfo()
>         if (c->x86_vendor == X86_VENDOR_CENTAUR && (c->x86 == 6) &&
>                 (c->x86_model == 7) || (c->x86_model == 8)) {
>
> That should probably be
>
>         if (c->x86_vendor == X86_VENDOR_CENTAUR && (c->x86 == 6) &&
>                 ((c->x86_model == 7) || (c->x86_model == 8))) {

Agreed, well spotted.
The same change also went into 2.4.10pre3 and needs the same fix there.

regards,

Dave.
-- 
| Dave Jones.        http://www.suse.de/~davej
| SuSE Labs

Re: Linux 2.4.9-ac6

[Andrea Arcangeli]

On Mon, Sep 03, 2001 at 02:50:47AM +0100, Alan Cox wrote:
> 2.4.9-ac6
> o	Add MODULE_LICENSE tags to telephony		(Robert Love)
> o	Add MODULE_LICENSE tags to drivers/video	(Robert Love)
> o	Add MODULE_LICENSE tags to zorro		(Robert Love)

what's the point of such information? If something I would understand to
specify the licence of a module when it's _not_ GPL.

Andrea

Re: Linux 2.4.9-ac6

[Alan Cox]

> what's the point of such information? If something I would understand to
> specify the licence of a module when it's _not_ GPL.

Because vendors will Im sure neglect to put non-GPL optional tags on their
code. While putting GPL tags on non GPL code is not going to make their
lawyers happy 8)

Alan

Re: Linux 2.4.9-ac6

[Keith Owens]

On Mon, 3 Sep 2001 15:05:29 +0200, 
Andrea Arcangeli <andrea@suse.de> wrote:
>On Mon, Sep 03, 2001 at 02:50:47AM +0100, Alan Cox wrote:
>> 2.4.9-ac6
>> o	Add MODULE_LICENSE tags to telephony		(Robert Love)
>> o	Add MODULE_LICENSE tags to drivers/video	(Robert Love)
>> o	Add MODULE_LICENSE tags to zorro		(Robert Love)
>
>what's the point of such information? If something I would understand to
>specify the licence of a module when it's _not_ GPL.

The next version of insmod will warn about modules with no
MODULE_LICENSE at all and inform about modules with proprietary
licences.  Both cases will mark the kenrel as tainted which will show
up on bug reports.

RE: Linux 2.4.9-ac6

[David Schwartz]

> On Mon, 3 Sep 2001 15:05:29 +0200,
> Andrea Arcangeli <andrea@suse.de> wrote:

> The next version of insmod will warn about modules with no
> MODULE_LICENSE at all and inform about modules with proprietary
> licences.  Both cases will mark the kenrel as tainted which will show
> up on bug reports.

	That really doesn't make sense. Nothing changes in the kernel or the module
based upon whether you have the source or not. What should logically taint
the kernel are modules that weren't compiled for that exact kernel version
or are otherwise mismatched.

	One can make the argument that the kernel is tainted if a module wasn't
compiled on that machine with that kernel version. One can make the argument
that the kernel is tainted if the module was compiled against different
configuration or header files. Once can make the argument that the kernel is
tainted if a module is loaded whose source isn't part of the general Linux
distribution. One can make all sorts of logical arguments about what taints
the kernel, but how can the license of a module taint the kernel?

	You can't even argue that if it's GPL, anyone can get the source to debug
it. The GPL does not require that the source code be made available to the
general public. Perhaps the kernel is tainted if that module wasn't built
from source on that machine?

	What's the logic here?!

	DS

Re: Linux 2.4.9-ac6

[Keith Owens]

On Tue, 4 Sep 2001 20:30:42 -0700, 
"David Schwartz" <davids@webmaster.com> wrote:
>	That really doesn't make sense. Nothing changes in the kernel or the module
>based upon whether you have the source or not. What should logically taint
>the kernel are modules that weren't compiled for that exact kernel version
>or are otherwise mismatched.

Bug reports when binary only modules have been loaded do not belong on
l-k, they have to go to the supplier.  AC wants to identify bug reports
that we can look at and ignore the ones that we cannot sensibly
investigate.  Any proprietary module loaded, at any time, means that
the bug report will almost certainly be ignored.

RE: Linux 2.4.9-ac6

[David Schwartz]

> On Tue, 4 Sep 2001 20:30:42 -0700,
> "David Schwartz" <davids@webmaster.com> wrote:

> >That really doesn't make sense. Nothing changes in the
> >kernel or the module
> >based upon whether you have the source or not. What should
> >logically taint
> >the kernel are modules that weren't compiled for that exact
> >kernel version
> >or are otherwise mismatched.
>
> Bug reports when binary only modules have been loaded do not belong on
> l-k, they have to go to the supplier.  AC wants to identify bug reports
> that we can look at and ignore the ones that we cannot sensibly
> investigate.  Any proprietary module loaded, at any time, means that
> the bug report will almost certainly be ignored.

	Yes, but even if the module is GPL'd, the module could still cost $1,000
and you're not entitled to the source if you didn't buy the module. If what
you want is "source code is available to the general public", then that can
be true or false for both GPL'd and non-GPL'd modules.

	If you make it 'must have been compiled from source on this machine" then
you know that at least this person has access to the source. If you make it
"source code must be publically avialable" then you know developers have
access to the  source. But the answer to both of these can be "yes" or "no"
for both GPL'd and non-GPL'd modules.

	What do you really want? Code that.

	DS

Re: Linux 2.4.9-ac6

[Alan Shutko]

"David Schwartz" <davids@webmaster.com> writes:

> Yes, but even if the module is GPL'd, the module could still cost
> $1,000 and you're not entitled to the source if you didn't buy the
> module.

OTOH, if you're getting a bug report from someone who has a GPLed
module, they can get the source and send it to you.

-- 
Alan Shutko <ats@acm.org> - In a variety of flavors!
Why do we have two eyes?  To watch 3-D movies with.

Re: Linux 2.4.9-ac6

[H. Peter Anvin]

Followup to:  <NOEJJDACGOHCKNCOGFOMOEBEDLAA.davids@webmaster.com>
By author:    "David Schwartz" <davids@webmaster.com>
In newsgroup: linux.dev.kernel
> 
> 	Yes, but even if the module is GPL'd, the module could still cost $1,000
> and you're not entitled to the source if you didn't buy the module. If what
> you want is "source code is available to the general public", then that can
> be true or false for both GPL'd and non-GPL'd modules.
> 

True enough, but if it's GPLd you are allowed to redistribute it free
of charge once you've bought it.

The main case of "unavailable GPL" code is probably modified GPL code
used within a company -- since only employees would have accessed to
the modified code, it simply would never get distributed.

	-hpa
-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
"Unix gives you enough rope to shoot yourself in the foot."
http://www.zytor.com/~hpa/puzzle.txt	<amsp@zytor.com>

Re: Linux 2.4.9-ac6

[Aaron Lehmann]

On Tue, Sep 04, 2001 at 10:16:15PM -0700, David Schwartz wrote:
> 	Yes, but even if the module is GPL'd, the module could still cost $1,000
> and you're not entitled to the source if you didn't buy the module. If what
> you want is "source code is available to the general public", then that can
> be true or false for both GPL'd and non-GPL'd modules.

WTF.

You understand that the GPL would allow the reporter to include the
module source, right? It also doesn't really matter if it doesn't come
with source because someone using the binary has a legal right to ask
for source.

RE: Linux 2.4.9-ac6

[David Schwartz]

> On Tue, Sep 04, 2001 at 10:16:15PM -0700, David Schwartz wrote:
> > 	Yes, but even if the module is GPL'd, the module could
> > still cost $1,000
> > and you're not entitled to the source if you didn't buy the
> > module. If what
> > you want is "source code is available to the general public",
> > then that can
> > be true or false for both GPL'd and non-GPL'd modules.

> WTF.
>
> You understand that the GPL would allow the reporter to include the
> module source, right? It also doesn't really matter if it doesn't come
> with source because someone using the binary has a legal right to ask
> for source.

	Sure, they have the legal right to ask for the source, but they might or
might not do so and might or might not make the source available to the
reporter.

	I think, perhaps, the logic should be that a module shouldn't taint the
kernel if:

	1) The user built the module from source on that machine, OR

	2) The module source is freely available without restriction

	But maybe I'm still not understanding what tainting is supposed to mean.
Note that both '1' and '2' are orthogonal to whether the module is GPL'd or
not.

	'2' would require a copyrighted tag, so that legal action could be taken
against those who misuse it (not that it's likely, considering it doesn't
really give any new capability and such deception would be rapidly
detected). I'm not sure if '1' is doable. Perhaps put a 'random' tag in the
config file and include that tag when the module is compiled. If the tags
match, then don't taint. This would catch config changes without module
rebuilds too.

	I think it's worth the effort to get this right to avoid sending the
message that if you use anything that's not GPL'd, the Linux community won't
help you. (Not that that's really what's happening, but it may look that
way.)

	DS

RE: Linux 2.4.9-ac6

[David Schwartz]

> 	I think, perhaps, the logic should be that a module
> shouldn't taint the kernel if:
>
> 	1) The user built the module from source on that machine, OR
>
> 	2) The module source is freely available without restriction

	I just realized two things. One, there's a strong argument that this should
be an AND, not an OR. Second, even if the code is freely available, it might
have been locally modified and the local modifications may not be easily
available (this can happen even if the code is GPL'd since source
distribution is only required if the compiled code is distributed). So the
stamping of the module source as freely available would have to have a
checksum or something like that. (Unless this is officially deemed to be No
Big Deal.)

	DS

Re: Linux 2.4.9-ac6

[Aaron Lehmann]

On Wed, Sep 05, 2001 at 12:03:39AM -0700, David Schwartz wrote:
> 	2) The module source is freely available without restriction

That disqualifies any GPL'd module.

RE: Linux 2.4.9-ac6

[David Schwartz]

> On Wed, Sep 05, 2001 at 12:03:39AM -0700, David Schwartz wrote:
> > 	2) The module source is freely available without restriction

> That disqualifies any GPL'd module.

	What I meant is that you can easily get the source and use it to debug. You
wouldn't have to register or pay a fee to obtain the source and no
restriction would prevent you from using to debug.

	DS

RE: Linux 2.4.9-ac6

[Justin Guyett]

On Wed, 5 Sep 2001, David Schwartz wrote:

> 	Sure, they have the legal right to ask for the source, but they might or
> might not do so and might or might not make the source available to the
> reporter.

If the reporters don't ask for the code, why should anyone help them solve
their problem by buying the module themselves and then requesting source
themselves?  Sure, it might happen, but I wouldn't expect it.

GPL *REQUIRES* the vendor make source available, it is not a "might or
might not" situation.  No scheme is perfect.  Of course even if your
kernel isn't tainted, if you're using a binary module that's GPL'd but the
company won't release source, it might as well be tained, but that's
an obvious case that really has no relevance here.

> 	I think, perhaps, the logic should be that a module shouldn't taint the
> kernel if:
>
> 	1) The user built the module from source on that machine, OR
>
> 	2) The module source is freely available without restriction
>
> 	But maybe I'm still not understanding what tainting is supposed to mean.
> Note that both '1' and '2' are orthogonal to whether the module is GPL'd or
> not.

It means nobody can help you because nobody outside the company that
produces the module has the source code.  1 is not necessarily valid
because bugs are present in code regardless of what architecture they're
run on, and regardless of whether the bug manifests itself.  It's
certainly more difficult to debug when the bug only shows up on other
people's systems, but such debugging isn't usually impossible.  The bigger
hurdle would be understanding the module well enough to spot problems even
if they did happen on your machine.

> 	'2' would require a copyrighted tag, so that legal action could be taken

GPL or GPL-compatible licenses are some of the few that have the necessary
qualities to allow debugging in a community.  You seem to be agreeing
somewhat, although the restriction of having to have bought the binary
isn't a problem in this case, so 2 is overly restrictive.

> 	I think it's worth the effort to get this right to avoid sending the
> message that if you use anything that's not GPL'd, the Linux community won't
> help you. (Not that that's really what's happening, but it may look that
> way.)

Absolutely.  Perhaps GPL-compatible licenses are not the only ones that
allow community debugging.  That's why the license is hopefully listed by
the module.  There's no force that will prevent someone from helping
someone else debug a module because the license matched "X", if that
someone can legally get the code and distribute patches.

The point isn't to be fascist, it's to point out that it's not usually
productive for people to go chasing down bugs that could be in code that
nobody has access to, either to notice or to fix the problem.  Again,
nothing's stopping someone from trying to debug a problem on a "tainted"
system.  Nobody's proposing a new kernel license that prohibits debugging
of a "tainted" kernel.


justin

RE: Linux 2.4.9-ac6

[Alex Bligh - linux-kernel]

>> 	I think, perhaps, the logic should be that a module
>> shouldn't taint the kernel if:
>>
>> 	1) The user built the module from source on that machine, OR
>>
>> 	2) The module source is freely available without restriction
>
> 	I just realized two things. One, there's a strong argument that this
> should be an AND, not an OR.

And as all distributions would fail (1) in initial form, all
distributions would result in tainted kernels. Is this the
intent?

--
Alex Bligh

RE: Linux 2.4.9-ac6

[David Schwartz]

> On Wed, 5 Sep 2001, David Schwartz wrote:

> > 	Sure, they have the legal right to ask for the source, but
> > they might or
> > might not do so and might or might not make the source available to the
> > reporter.

> If the reporters don't ask for the code, why should anyone help them solve
> their problem by buying the module themselves and then requesting source
> themselves?  Sure, it might happen, but I wouldn't expect it.

	Exactly! So *force* the submitter to actually *have* the source code to the
module. It doesn't matter whether the module is GPL or not, it just matters
whether the submitter compiled the module from source code which he has.

> GPL *REQUIRES* the vendor make source available, it is not a "might or
> might not" situation.

	Under some circumstances, yes, but not all. Suppose, for example, the
module isn't distributed. Suppose the written offer for source expired.

> No scheme is perfect. Of course even if your
> kernel isn't tainted, if you're using a binary module that's GPL'd but the
> company won't release source, it might as well be tained, but that's
> an obvious case that really has no relevance here.

	Right. So, as I suggested, consider the kernel tainted if there are any
modules loaded that weren't compiled from source on that box with that
kernel version and headers. Suggest the user recompile his modules,
reproduce the problem, and submit a bug report with the source or URL to any
modules. Problem solved, no licensing prejudice involved.

> > 	I think, perhaps, the logic should be that a module
> > shouldn't taint the
> > kernel if:
> >
> > 	1) The user built the module from source on that machine, OR
> >
> > 	2) The module source is freely available without restriction
> >
> > 	But maybe I'm still not understanding what tainting is
> > supposed to mean.
> > Note that both '1' and '2' are orthogonal to whether the module
> > is GPL'd or
> > not.

> It means nobody can help you because nobody outside the company that
> produces the module has the source code.  1 is not necessarily valid
> because bugs are present in code regardless of what architecture they're
> run on, and regardless of whether the bug manifests itself.  It's
> certainly more difficult to debug when the bug only shows up on other
> people's systems, but such debugging isn't usually impossible.  The bigger
> hurdle would be understanding the module well enough to spot problems even
> if they did happen on your machine.

	So tell the user that he must be willing to provide the source code to any
modules in order to have his bug report honored. Another possibility would
be a system to tag modules with a URL from which the source could be freely
downloaded and include that in the bug report. If that tag was not present,
then taint.

> > 	'2' would require a copyrighted tag, so that legal action
> > could be taken

> GPL or GPL-compatible licenses are some of the few that have the necessary
> qualities to allow debugging in a community.  You seem to be agreeing
> somewhat, although the restriction of having to have bought the binary
> isn't a problem in this case, so 2 is overly restrictive.

	If '2' isn't met, then tell the user that he *must* be willing to make the
source available to anyone who wishes to help debug. If he isn't, then he
won't get help.

> > 	I think it's worth the effort to get this right to avoid sending the
> > message that if you use anything that's not GPL'd, the Linux
> > community won't
> > help you. (Not that that's really what's happening, but it may look that
> > way.)

> Absolutely.  Perhaps GPL-compatible licenses are not the only ones that
> allow community debugging.  That's why the license is hopefully listed by
> the module.  There's no force that will prevent someone from helping
> someone else debug a module because the license matched "X", if that
> someone can legally get the code and distribute patches.

> The point isn't to be fascist, it's to point out that it's not usually
> productive for people to go chasing down bugs that could be in code that
> nobody has access to, either to notice or to fix the problem.  Again,
> nothing's stopping someone from trying to debug a problem on a "tainted"
> system.  Nobody's proposing a new kernel license that prohibits debugging
> of a "tainted" kernel.

	I hope I'm not overreacting. It just seems that tainting based upon
licensing doesn't do what's really wanted, which is to make sure that the
user has the source and those who wish to debug can also get the source. The
first part of that is really easy to check directly (and you catch stupidity
due to mismatched versions, SMP versus UP, and so on). I think it makes
sense to 'force' the user to replicate the bug after having cleanly compiled
the module. This ensures that at least he has the source.

	DS

RE: Linux 2.4.9-ac6

[David Schwartz]

> >> 	I think, perhaps, the logic should be that a module
> >> shouldn't taint the kernel if:

> >> 	1) The user built the module from source on that machine, OR

> >> 	2) The module source is freely available without restriction

> > 	I just realized two things. One, there's a strong argument that this
> > should be an AND, not an OR.

> And as all distributions would fail (1) in initial form, all
> distributions would result in tainted kernels. Is this the
> intent?

	They wouldn't taint because the kernel signature would match the module
signatures. I provided more detail on one possible way this scheme would
work and it's not quite as simple as the summary above suggests.

	Basically, when you compile (or install) the kernel, a random 'signature'
goes in. When you compile a module, the signature goes in too. You can then
compare the module's signature to the kernel's signature to ensure they were
compiled as a unit. Unfortunately, this doesn't ensure that the user has the
source.

	I suppose, if the module source were freely available, then '2' would
apply. If you keep it as an OR, then distributions wouldn't taint out of the
box unless they included modules whose source distribution was limited. I
think this is what you want.

	DS

RE: Linux 2.4.9-ac6

[Justin Guyett]

On Wed, 5 Sep 2001, David Schwartz wrote:

<snip>

So basically this is all obvious.  Nobody's going to help someone with a
million-dollar GPL module if they don't provide the code.  People will
help others with non-GPL modules if they are interested in helping, can
legally get the code, and can legally distribute patches.  Nobody's going
to be an idiot and stick rigidly by a tainting scheme when someone
obviously has an open-source module that's not GPL, and needs help.
That's all I was trying to say.  I don't think everyone's going to
ignore tainted bug reports universally, and there's nothing requiring that
to happen.  It makes sense to read a bit more into a bug report anyway;
Ignoring filesystem corruption because the bug report was tainted by an
nvidia driver or a 3rd-party serial driver isn't going to help anyone.

That's all in the interpretation of the tainted report, not the fact that
it's tainted.  There's also reporter-end probably-undebuggable report
elimination.  If two people a year don't report "X Crashed ... and I use
nvdriver", that's a reason to have a taint flag, even if everyone ignored
the flag, even if it was set, in bug reports that were posted, right?


justin

Re: Linux 2.4.9-ac6

[Alan Cox]

> 	I hope I'm not overreacting. It just seems that tainting based upon
> licensing doesn't do what's really wanted, which is to make sure that the
> user has the source and those who wish to debug can also get the source. The
> first part of that is really easy to check directly (and you catch stupidity
> due to mismatched versions, SMP versus UP, and so on). I think it makes
> sense to 'force' the user to replicate the bug after having cleanly compiled
> the module. This ensures that at least he has the source.

The goal is very simple. It is to make the information available. What
people choose to do with that information and such bug reports is up to
them. Similarly the policy is in user space modutils.

Unfortunately I get so many bug reports caused by the nvidia modules and
people lying when asked if they have them loaded that some kind of action
has to occur, otherwise I'm going to have to stop reading bug reports from
anyone I don't know personally.

I am not prepared to be an unpaid Nvidia support goon. 

Alan

Re: Linux 2.4.9-ac6

[Alan Cox]

> based upon whether you have the source or not. What should logically taint
> the kernel are modules that weren't compiled for that exact kernel version
> or are otherwise mismatched.

Setting a flag for the insmod -f required case as well is an extremely good
idea. This is entirely about making information available nothing else and
your suggestion there is a good one.

Alan

Re: Linux 2.4.9-ac6

[Arjan van de Ven]

Alan Cox wrote:
> 
> > based upon whether you have the source or not. What should logically taint
> > the kernel are modules that weren't compiled for that exact kernel version
> > or are otherwise mismatched.
> 
> Setting a flag for the insmod -f required case as well is an extremely good
> idea. This is entirely about making information available nothing else and
> your suggestion there is a good one.

How about making the "tainted" field a bitmask ?
eg bit 0 --> non GPL/BSD module
   bit 1 --> insmod -f

Greetings,
    Arjan van de Ven

Re: Linux 2.4.9-ac6

[Jan-Benedict Glaw]

On Wed, 2001-09-05 13:16:58 +0100, Arjan van de Ven <arjanv@redhat.com>
wrote in message <3B9617BA.F771914E@redhat.com>:
> Alan Cox wrote:
> > > based upon whether you have the source or not. What should logically taint
> > > the kernel are modules that weren't compiled for that exact kernel version
> > > or are otherwise mismatched.
> > Setting a flag for the insmod -f required case as well is an extremely good
> > idea. This is entirely about making information available nothing else and
> > your suggestion there is a good one.
> 
> How about making the "tainted" field a bitmask ?
> eg bit 0 --> non GPL/BSD module
>    bit 1 --> insmod -f

Basically, I don't like that idea. 'insmod -f' should only be
required if the module in question is some kind of a commercial (TM)
module. Setting a "GPL/BSD" flag might be somewhat interesting
(but enlarges needed on-disk-space), but I don't like to help
those commercials to ease the use of their broken, binary-only
modules.

MfG, JBG

-- 
Jan-Benedict Glaw . jbglaw@lug-owl.de . +49-172-7608481

Re: Linux 2.4.9-ac6

[christophe barbé]

Would it not be possible with your scheme to package a closed source driver
in an open source wrapper driver and then defeat your tainting technique.

Is it legally possible to copyright a kind of magic number with a copyright
allowing only it's used in open & public source driver ?

Christophe 


Le mer, 05 sep 2001 11:57:17, David Schwartz a écrit :
> 
> > >> 	I think, perhaps, the logic should be that a module
> > >> shouldn't taint the kernel if:
> 
> > >> 	1) The user built the module from source on that machine, OR
> 
> > >> 	2) The module source is freely available without restriction
> 
> > > 	I just realized two things. One, there's a strong argument
> that this
> > > should be an AND, not an OR.
> 
> > And as all distributions would fail (1) in initial form, all
> > distributions would result in tainted kernels. Is this the
> > intent?
> 
> 	They wouldn't taint because the kernel signature would match the
> module
> signatures. I provided more detail on one possible way this scheme would
> work and it's not quite as simple as the summary above suggests.
> 
> 	Basically, when you compile (or install) the kernel, a random
> 'signature'
> goes in. When you compile a module, the signature goes in too. You can
> then
> compare the module's signature to the kernel's signature to ensure they
> were
> compiled as a unit. Unfortunately, this doesn't ensure that the user has
> the
> source.
> 
> 	I suppose, if the module source were freely available, then '2'
> would
> apply. If you keep it as an OR, then distributions wouldn't taint out of
> the
> box unless they included modules whose source distribution was limited. I
> think this is what you want.
> 
> 	DS
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel"
> in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 
> 
-- 
Christophe Barbé
Software Engineer - christophe.barbe@lineo.fr
Lineo France - Lineo High Availability Group
42-46, rue Médéric - 92110 Clichy - France
phone (33).1.41.40.02.12 - fax (33).1.41.40.02.01
http://www.lineo.com

Re: Linux 2.4.9-ac6

[Alan Cox]

> Would it not be possible with your scheme to package a closed source dr=
> iver
> in an open source wrapper driver and then defeat your tainting techniqu=
> e.
> 
> Is it legally possible to copyright a kind of magic number with a copyr=
> ight
> allowing only it's used in open & public source driver ?

Oh you can certainly defeat it. That doesn't actually bother me. Firstly
nobody at Nvidia and friends is trying to mislead us, its their end users
who are. Secondly your average corporate lawyers get upset about putting
in license tags that are not true in case a court holds them to the tags,
and thirdly since its arguably digital rights management they might face
five years in jail in the USA for doing so 8)

Alan

Re: Linux 2.4.9-ac6

[Alex Bligh - linux-kernel]

> Would it not be possible with your scheme to package a closed source
> driver in an open source wrapper driver and then defeat your tainting
> technique.

It would also be theoretically possible for an evil driver merchant
to twiddle the flag back via /dev/kmem (for instance). Or load the
module by manipulation of /dev/kmem. Or for the bug-reporting user
to patch their kernel so that the flag never got set and hence
disguise the presence of an nvidia driver (etc.) in a misguided
attempt to wangle support out of Alan et al.

However, I understood the point of the exercize to be a first pass
hueristic to flag bug reports from systems running modules for
which Alan and others haven't got, and can't get the source. It's
not going to be perfect (see above), but equally doesn't need to be.
I'm sure users do all sorts of other 'well don't do that, then'
stuff which wastes the time of those reading bug reports.

--
Alex Bligh

Re: Linux 2.4.9-ac6

[christophe barbé]

Le mer, 05 sep 2001 15:14:09, Alex Bligh - linux-kernel a écrit :
> > Would it not be possible with your scheme to package a closed source
> > driver in an open source wrapper driver and then defeat your tainting
> > technique.
> 
> It would also be theoretically possible for an evil driver merchant
> to twiddle the flag back via /dev/kmem (for instance). Or load the
> module by manipulation of /dev/kmem. Or for the bug-reporting user
> to patch their kernel so that the flag never got set and hence
> disguise the presence of an nvidia driver (etc.) in a misguided
> attempt to wangle support out of Alan et al.
> 
> However, I understood the point of the exercize to be a first pass
> hueristic to flag bug reports from systems running modules for
> which Alan and others haven't got, and can't get the source. It's
> not going to be perfect (see above), but equally doesn't need to be.
> I'm sure users do all sorts of other 'well don't do that, then'
> stuff which wastes the time of those reading bug reports.

Yes I agree that's not easy and certainly not the goal to avoid this kind
of thing.

Btw I was thinking about a real case: I use in my laptop the lucent driver
for their winmodem chipset. This driver is closed source but we use it
relinked with proper opensource code. This avoid the use of 'insmod -f' and
most of the bug (caused by missing symbols) but you can not trust the
resulting module.

Christophe

> --
> Alex Bligh
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel"
> in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 
> 
-- 
Christophe Barbé
Software Engineer - christophe.barbe@lineo.fr
Lineo France - Lineo High Availability Group
42-46, rue Médéric - 92110 Clichy - France
phone (33).1.41.40.02.12 - fax (33).1.41.40.02.01
http://www.lineo.com

Re: Linux 2.4.9-ac6

[Henning P. Schmiedehausen]

christophe =?iso-8859-1?Q?barb=E9?= <christophe.barbe@lineo.fr> writes:

>Would it not be possible with your scheme to package a closed source driver
>in an open source wrapper driver and then defeat your tainting technique.

>Is it legally possible to copyright a kind of magic number with a copyright
>allowing only it's used in open & public source driver ?

Congratulations, you've just invented the Microsoft HW Labs
certification procedure with module signing. ;-)

I can really see it:

# insmod <module>
This module <module> is not signed with the Linus Torvalds Hardware
Certification Labs key.

Do you want to

a) insert the module anyway (Errors and Crashes from your kernel will
not be processed by major kernel developers
b) not insert the module
c) search for an alternative driver that is open source, download it,
   compile it and use that instead of <module>
d) download a skeleton driver to write your own driver
e) install Microsoft Windows XP
f) exit

-> f
# 

	Regards
		Henning

-- 
Dipl.-Inf. (Univ.) Henning P. Schmiedehausen       -- Geschaeftsfuehrer
INTERMETA - Gesellschaft fuer Mehrwertdienste mbH     hps@intermeta.de

Am Schwabachgrund 22  Fon.: 09131 / 50654-0   info@intermeta.de
D-91054 Buckenhof     Fax.: 09131 / 50654-20   

RE: Linux 2.4.9-ac6

[David Schwartz]

> "David Schwartz" <davids@webmaster.com> writes:

> > Yes, but even if the module is GPL'd, the module could still cost
> > $1,000 and you're not entitled to the source if you didn't buy the
> > module.

> OTOH, if you're getting a bug report from someone who has a GPLed
> module, they can get the source and send it to you.

	Perhaps, but:

	1) They may not have it or know the procedure to get it.

	2) The offer for the source may have expired.

	3) This is true whether or not the module is GPLed, they may have the
source or they may not.

	4) Many other licenses may result in publically available source and may
not, just like the GPL.

	If you want to make sure the user has the source from which the module was
built, that can be checked for. (Though it's not simple, unfortunately.)

	What you want to do is make sure the user understands that he had better be
able to produce the source for any kernel modules loaded at the time of the
problem. If not, he needs to reproduce the bug on a system without the
relevant modules.

	I think a tag for 'source code is available to anyone for debugging
purposes, no questions asked' is more what's wanted.

	DS

RE: Linux 2.4.9-ac6

[David Schwartz]

> Unfortunately I get so many bug reports caused by the nvidia modules and
> people lying when asked if they have them loaded that some kind of action
> has to occur, otherwise I'm going to have to stop reading bug reports from
> anyone I don't know personally.
>
> I am not prepared to be an unpaid Nvidia support goon.
>
> Alan

	Well, now you have the same problem everyone else has with open source. You
can't protect yourself from the user by changing the source code because the
user has the source code. In any event, coding Linux to prevent the user
from doing something he wants to do, no matter how wrong you think that
something is, just doesn't feel right.

	I guess it's just sad that this is the issue. What stops someone from
editing the taint lines out of the bug report? Are you going to have the
kernel checksum them? *sigh*

	DS

Re: Linux 2.4.9-ac6

[Alan Cox]

> 	Well, now you have the same problem everyone else has with open source. You
> can't protect yourself from the user by changing the source code because the
> user has the source code. In any event, coding Linux to prevent the user

Oh thats never been a problem. I've never had trouble with people hacking
source and denying it, nor with helping them.

> 	I guess it's just sad that this is the issue. What stops someone from
> editing the taint lines out of the bug report? Are you going to have the
> kernel checksum them? *sigh*

The people using the Nvdriver and causing the problem don't have the skills
to do it. 

Alan

RE: Linux 2.4.9-ac6 (really about tainting)

[David Schwartz]

	I fear that out of fear that someone might make a mountain out of a
molehill, I've made a mountain out of a molehill. I hope at least that some
of my suggestions prove useful. Sorry about any heat that got mixed in with
the light.

	DS