linux-2.4.27 released

linux-2.4.27 released

[Marcelo Tosatti]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 44701 bytes --]

final:

- 2.4.27-rc6 was released as 2.4.27 with no changes.

Here is a list of the most important security issues fixed by this release:

CAN-2004-0495 (Al Viro sparse fixes)
CAN-2004-0497 (users could modify group ID of arbitrary files on the system)
CAN-2004-0535 (e1000 minor info leak)
CAN-2004-0685 (backported Conectiva usb sparse fixes)
CAN-2004-0415 (file offset pointer handling race)
CAN-2004-0565 (information leak ia64)


Summary of changes from v2.4.27-rc5 to v2.4.27-rc6
============================================

Geert Uytterhoeven:
  o Fix net/atm/br2684.c file offset patch

Marcelo Tosatti:
  o Changed EXTRAVERSION to -rc6

Mikael Pettersson:
  o drivers/macintosh/nvram.c typo


Summary of changes from v2.4.27-rc4 to v2.4.27-rc5
============================================

Adrian Bunk:
  o [NET]: Update CONFIG_NET_SCH_NETEM Configure.help entry

Chris Wright:
  o ethtool_get_regs copy right number of bytes to user

Eric Sandeen:
  o [XFS] Don't lock down user pages when doing direct IO; this can lead to trouble (double-locking zero page, etc).

Harald Welte:
  o [NETFILTER]: ipt_ULOG fix for packet delay
  o [NETFILTER]: Fix broken debug assertion

Marcel Holtmann:
  o [Bluetooth] Fix kernel panic when device config fails
  o [Bluetooth] Replace BCSP retransmitting message with BT_DBG
  o [Bluetooth] Fix resetting to default filters
  o [Bluetooth] Send HCI_Reset for ISSC USB dongles

Marcelo Tosatti:
  o Revert DVD-RW write support for now: axboe@suse.de|ChangeSet|20040606235035|46544
  o Cset exclude: axboe@suse.de|ChangeSet|20040607195639|57919
  o Remove mm/page_alloc.c debugging
  o Al Viro and others: Fix file offset handling races in several drivers
  o Changed EXTRAVERSION to -rc5
  o update-i386-defconfig.patch

Nathan Scott:
  o [XFS] Fix data loss problem - we no longer update i_size anywhere without holding i_sem for 2.4 as well.
  o [XFS] Fix diotest4 test case issues with direct reads in XFS

Ralf Bächle:
  o Fix non-use of HZ in 6pack.c

Stephen Hemminger:
  o bridge fix
  o [TCP]: Bic tcp congestion calculation timestamp
  o [PKT_SCHED]: netem limit not returned correctly


Summary of changes from v2.4.27-rc3 to v2.4.27-rc4
============================================

<ajgrothe:yahoo.com>:
  o [CRYPTO]: Add TEA and XTEA algorithms

<jaap.keuter:xs4all.nl>:
  o [IPV4]: Calculate default broadcast even when using SIOCSIGNETMASK

<mjc:redhat.com>:
  o USB: more sparse fixes

<tomd:csds.uidaho.edu>:
  o [CRYPTO]: Set CRYPTO_TFM_RES_BAD_KEY_LEN in twofish

Adrian Bunk:
  o [IPV4]: Remove no longer available URL
  o cmpci oops on rmmod + fix

Anton Blanchard:
  o [TG3]: Missing rmb() in rx processing

Arun Sharma:
  o ia64: tighten FPH state context switch check

Dave Kleikamp:
  o JFS: Error path released metadata page it shouldn't have
  o JFS: Updated field isn't always written to disk during truncate
  o JFS: Protect active_ag with a spinlock
  o JFS: prevent concurrent calls to txCommit on the imap inode
  o JFS: Check for dmap corruption before using leafidx
  o JFS: jfs_dmap build fix

David S. Miller:
  o [TG3]: Always do 4gb tx dma test, and fix the test
  o [TG3]: Fibre PHY fixes from Sun
  o [TG3]: Update driver version and reldate
  o [TG3]: Delay both before and after PCI cfg space readback after reset
  o [TG3]: Bump driver version and reldate
  o [TG3]: Update reldate to match 2.6.x sources
  o [IPV4]: Make raw sockets behave like udp wrt. MSG_TRUNC
  o [ATM]: Update Marko Kiiskila's email address
  o [PKT_SCHED]: sch_netem.c needs linux/init.h
  o [CRYPTO]: No MODULE_ALIAS in 2.4.x

Geert Uytterhoeven:
  o M68k ifpsp060
  o M68k 68060 errata I14
  o M68k Maintainership

Herbert Xu:
  o [CRYPTO]: Fix stack overrun in crypt()

Jochen Hein:
  o Update Jochen CREDITS entry

Karsten Keil:
  o I4L: Fix IRQ-sharing lockup in nj_s

Marcel Holtmann:
  o [Bluetooth] Respond to L2CAP info requests
  o [Bluetooth] Add support for another ALPS module
  o [Bluetooth] Use a signed integer for the RSSI value

Marcelo Tosatti:
  o USB: fix endless resubmit in auerswald (Wolfgang Mues)
  o Changed EXTRAVERSION to -rc4

Masanari Iida:
  o Fix harmless typo in drivers/char/sysrq.c

Mikael Pettersson:
  o cardbus.c pointer truncation bug on 64-bitters

Patrick McHardy:
  o [IPV4/IPV6]: Add myself to MAINTAINERS

Pete Zaitcev:
  o USB: update unusual_devs.h
  o USB: GET_ID from nonzero interface (errandir_news@mph.eclipse.co.uk)
  o USB: add free_len=0 initialization to ipaq.c (Ganesh Varadaraja)
  o USB: correct dbg() arguments in pl2303 (Phil Dibowitz)
  o USB: missing rcomplete=0 in printer.c (David Woodhouse)

Petr Vandrovec:
  o [VLAN]: Do not access released memory

Samuel Thibault:
  o [UDP]: Return true length if user specifies MSG_TRUNC

Stephen Hemminger:
  o [PKT_SCHED]: Update to network emulation QOS scheduler
  o [PKT_SCHED]: One small netem fixes
  o [BRIDGE]: Fix assertion failure in 2.4.27-rc3
  o [PKT_SCHED]: netem update for 2.4

Stéphane Eranian:
  o ia64: fix info in /proc/pal/*/bus_info
  o ia64: fix perfmon buffer init


Summary of changes from v2.4.27-rc2 to v2.4.27-rc3
============================================

Adrian Bunk:
  o add missing USB Gadget Configure.help entries

Christoph Hellwig:
  o [NETLINK]: Fix NLMSG_OK/RTA_OK length checking

David S. Miller:
  o [NET]: Fix SO_{RCV,SND}TIMEO getsockopt handling

Jan-Benedict Glaw:
  o [NETFILTER]: ip_fw_compat_masq.c needs net/ip.h

Marcelo Tosatti:
  o Thomas Biege: Fix missing DAC check on sys_chown
  o Changed EXTRAVERSION to -rc3
  o Add missing bracket to inode_change_ok() fix

Stephen Hemminger:
  o [BRIDGE]: Bridge STP message age

Tom Rini:
  o Li Yang: PQII FCC Ethernet driver: transmit buffer leak


Summary of changes from v2.4.27-rc1 to v2.4.27-rc2
============================================

<glen:imodulo.com>:
  o Fix comment typo in nForce2 C1halt fixup

<skraw:ithnet.com>:
  o orinoco_pci.c new device 0x3872

Adrian Bunk:
  o add ATM_FORE200E_USE_TASKLET Configure.help entry
  o add missing E1000_NAPI Configure.help text
  o add SATA Configure.help texts

Christoph Hellwig:
  o [XFS] Don't leak locked pages on readahead failure

Dave Kleikamp:
  o jfs warning fix

David Brownell:
  o usb gadget API updates (1/5)
  o usb gadget zero updates (2/5)
  o usb file storage gadget updates (3/5)
  o usb ethernet+RNDIS gadget updates (4/5)
  o usb gadget build updates (5/5)

David S. Miller:
  o [PKT_SCHED]: Do not check netif_queue_stopped() in dequeue ops, races with driver
  o [NET]: Check __dev_get_by_name() return value in eql.c

David Stevens:
  o [IPV6]: Handle user asking for any device in mcast calls

Dean Roehrich:
  o [XFS] Change things to use new version of xfs_dm_init/xfs_dm_exit
  o [XFS] Fix non-dmapi build

Eric Sandeen:
  o [XFS] Fix overflow in mapping test at offsets of 2^63-1 bytes

Geert Uytterhoeven:
  o affs remount
  o Mac Sonic Ethernet fixes
  o Mac IOP: Fix bug found by Opera

Harald Welte:
  o [NETFILTER]: Fix non-existant config option
  o [NETFILTER]: Complain when broken ipt_owner options are used on SMP
  o [NETFILTER]: Change permissions of /proc/net/ip_conntrack to 0440
  o [NETFILTER]: skip internal targets in iptables proc listing
  o [NETFILTER]: Fix inverted matching in ipt_helper
  o [NETFILTER]: 'any' matching in ipt_helper
  o [NETFILTER]: Don't reroute on nfmark change in mangle table when routing by nfmark is not enabled
  o [NETFILTER]: Fix truncated fragment check in ipt_unclean
  o [NETFILTER]: Remove broken check for cleared IP_DF flag on fragments in ipt_unclean
  o [NETFILTER]: Add new function 'nf_reset' to reset netfilter related skb-fields

Hideaki Yoshifuji:
  o [IPV6]: Fix autoconf description in ip-sysctl.txt

Jeb J. Cramer:
  o e1000 management reset fix

Jeff Garzik:
  o [netdrvr] fix ethtool_ops design bug, sync with 2.6.x ethtool_ops code
  o Rename 'carmel' block driver to 'sx8'

Manfred Spraul:
  o yenta: Add override_bios flag to ignore BIOS resource allocation

Marcel Holtmann:
  o [Bluetooth] Kill duplicate includes
  o [Bluetooth] Update help entries
  o [Bluetooth] Allocate protocol number for HIDP support
  o [Bluetooth] Add quirk for broken RTX Telecom based dongles
  o [Bluetooth] Fix connection creation error handling

Marcelo Tosatti:
  o Geert Uytterhoeven: Fix warnings and cleanup debug_page() addded in -rc1
  o Al Viro sparse fixes: decnet user pointer dereference
  o Al Viro sparse fixes: mpu401 user pointer dereference
  o Al Viro sparse fixes: msnd user pointer dereference & assorted fixes
  o Al Viro sparse fixes: pss user pointer dereference
  o Al Viro sparse fixes: aironet
  o Al Viro sparse fixes: asus_acpi user pointer dereference
  o Changed EXTRAVERSION to -rc2

Matt Domsch:
  o edd.c display %u, remove REPORT_URL

Mikael Pettersson:
  o i386 and x86_64 ACPI mpparse timer bug

Nathan Scott:
  o [XFS] No longer hold the BKL for the entire ioctl operation, its not needed here.
  o [XFS] Remove a couple of redundant NULL parent inode pointer checks
  o [XFS] Fix xfs_lowbit64, it mishandled zero in the high bits
  o [XFS] sparse: fix uses of macros before their definitions, etc
  o [XFS] Ensure buffers that map to unwritten extents are only submitted when properly setup.
  o [XFS] Sanitise the ACL initialisation macros
  o [XFS] Remove unused MAC macros, never needed on Linux
  o [XFS] Remove the one remaining, broken use of XFS_WRITEIO_LOG and sanitize direct IO map blocks call.
  o [XFS] Fix flags argument to xfs_incore call on attr removal
  o [XFS] Fix a race condition in the undo-delayed-write buffer routine
  o [XFS] Fix up memory allocators to be more resilient
  o [XFS] Fix up error handling on inode shrink register
  o [XFS] Fix up memory reclaim interfaces for 2.4 kernels

Pete Zaitcev:
  o Lonnie Mendez: Remove USB extra #include

Russell Cattelan:
  o [XFS] Fix for NFS+XFS data corruption problem

Scott Feldman:
  o e1000: fix napi crash on ifdown during traffic

Stephen Hemminger:
  o [PKT_SCHED]: Delay scheduler enqueue always succeeds
  o [PKT_SCHED]: Delay scheduler should retry if requeue fails
  o [PKT_SCHED]: Add loss option to network delay scheduler
  o [BRIDGE]: Backport of API checking
  o [BRIDGE]: Elimintate br_ioctl_mutex


Summary of changes from v2.4.27-pre6 to v2.4.27-rc1
============================================

<andreas:xss.co.at>:
  o fix hotplug Config.in xconfig breakage

Andi Kleen:
  o Undo mistaken hunks in previous x86-64 MCE change
  o Add missing include to x86-64 bluesmoke.c

Chris Wedgwood:
  o stat nlink resolution fix

David S. Miller:
  o Cset exclude: kaber@trash.net|ChangeSet|20040529193918|20643

David Stevens:
  o [IPV4]: Fix interface selection in multicast sockops

Eyal Lebedinsky:
  o Fix USB visor.c compilation error

Hideaki Yoshifuji:
  o [IPV6]: UDPv6 checksum
  o [IPV6]: UDPv6: Use udpv6_queue_rcv_skb()
  o [IPV6]: Missing include in ip6_tables.c

Karol Kozimor:
  o acpi4asus trivial sync with 2.6 (Karol 'sziwan' Kozimor)

Len Brown:
  o [ACPI] PCI bus numbering workaround for ServerWorks from David Shaohua Li http://bugzilla.kernel.org/show_bug.cgi?id=1662
  o [ACPI] fix passive cooling mode indicator (Luming Yu) http://bugzilla.kernel.org/show_bug.cgi?id=1770
  o [ACPI] avoid spurious interrupts on VIA http://bugzilla.kernel.org/show_bug.cgi?id=2243
  o [ACPI] fix 2.4.27-pre3 IRQ override regression due to dynamically allocated mp_irqs[].
  o [ACPI] handle SCI override to nth IOAPIC http://bugzilla.kernel.org/show_bug.cgi?id=2835

Marcelo Tosatti:
  o Add missing struct definition of rwsem race fixes
  o Changed EXTRAVERSION to -rc1
  o journal_try_to_free_buffers(): Add debug print in case of bh list corruption

Mikulas Patocka:
  o HPFS fixes

Pete Zaitcev:
  o USB: Fix jumpshot's capacity


Summary of changes from v2.4.27-pre5 to v2.4.27-pre6
============================================

<andrej.filipcic:ijs.si>:
  o USB: pl2303 & input overruns

<chaapala:cisco.com>:
  o [CRYPTO]: Fix digest.c kmapping sg entries > page in length

<hadi:zynx.com>:
  o [NETFILTER]: Small interface cleanup for {ipt,ip6t,arpt}_find_target

<jan:ccsinfo.com>:
  o USB ftdi device ids for 2.4

<jhh:lucent.com>:
  o [SCTP] Fix to not setup a new association if the endpoint is in SHUTDOWN_ACK_SENT state and recognizes that the peer has restarted. 
  o [SCTP] Fix to not start a new association on a 1-many style sendmsg() with MSG_EOF/MSG_ABORT flag and no data.

<juhl-lkml:dif.dk>:
  o [NET]: Remove junk from packet_mmap.txt

<kumar.gala:freescale.com>:
  o Simple build fix for PPC 826x

<martin.lubich:gmx.at>:
  o USB: add Clie TH55 Support in visor kernel module

<nickpiggin:yahoo.com.au>:
  o rwsem race fixes backported from 2.6

<raven:themaw.net>:
  o minor autofs4 fs/namei.c fix, URL update

<shurick:sectorb.msk.ru>:
  o i2c-matroxfb and i2c initialization order

Andi Kleen:
  o Fix argument parsing in x86-64 machine check handler
  o Fix LDT/TSS limit on x86-64
  o Fix compilation without CONFIG_SWIOTLB on x86-64
  o Fix boot loader warnings on x86-64
  o Add missing memory clobber to i386
  o Fix K8 machine check decoding
  o Fix FPU delayed exceptions on x86-64 too

Andreas Dilger:
  o [IPV4]: Fix bug in arp_tables.c fix

Bartlomiej Zolnierkiewicz:
  o clarify help entries for Promise IDE drivers a bit
  o ide: add new nForce IDE/SATA device IDs to amd74xx.c

Ben Collins:
  o video1394: Bugfix for low res format7 images
  o sbp2: Default sbp2_serialize_io to 1
  o pcilynx: Disable LCtrl bit with IRM contender

Dave Kleikamp:
  o JFS: Handle out of space errors more gracefully
  o JFS: Better RAS when btstack is overrun
  o JFS: Don't allow reading beyond the inode map's EOF
  o JFS: Fix compilation error

David Brownell:
  o EHCI fixes (byteswap, BIOS)
  o EHCI fixes (byteswap, BIOS)

David S. Miller:
  o [TG3]: Use HOST TXDs always
  o [IPV4]: Fix unaligned accesses in arp_tables.c
  o [TG3]: Chip support update and a power-save bug fix
  o [TG3]: Update driver version and reldate
  o [SPARC]: Report si_addr in SIGINFO more accurately

Greg Kroah-Hartman:
  o USB: fix empty write issue in pl2303 driver

Harald Welte:
  o [NETFILTER]: Don't assign new helper after NAT when there are already expectations present

Herbert Xu:
  o Fix /proc/ide/hpt366 crash

Hideaki Yoshifuji:
  o [UDPv4]: Pass correct socket to ip_mc_sf_allow

Ivan Kokshaysky:
  o Alpha: don't put IDE disks in standby mode on halt
  o Alpha: fix PCI bridge swizzle on takara and eiger

Jamal Hadi Salim:
  o [NET]: Add ARPHRD_NONE and use it in tun driver

Jens Axboe:
  o DVD-RW write support
  o cdrom hardware defect mgt header length

Marcelo Tosatti:
  o Cset exclude: marcelo@logos.cnet|ChangeSet|20040606231521|43962
  o Thomas Winischhofer: sisfb update
  o Alexander Nyberg/Andi/Sergey: Fix x86 "clear_cpu()" macro
  o Changed EXTRAVERSION to -pre6
  o Michael Reinelt: Add support for NetMos 9835 serial cards
  o John Carlson: Remove bogus ";" from USB gadgets's usb_descriptor_fillbuf

Mike Miller:
  o cciss update

Neil Brown:
  o Fix hard-to-hit BUG in raid5 resync code

Nuno Monteiro:
  o Fix rwsem-fix typo
  o Complete rwsem typo fix

Pete Zaitcev:
  o Fix USB serial race
  o USB: Update mct_u232
  o USB: unusual_dev.h add Sony Handycam HC-85

Sridhar Samudrala:
  o [SCTP] Fix the use of cached non-zero vtag in an INIT that is resent after a stale cookie error.
  o [SCTP] Fix missing VTAG validation on certain incoming packets
  o [SCTP] Fix to wakeup blocking connect() after max INIT retries failed
  o [SCTP] Fix poll() on a 1-1 style socket so that it returns when the association is aborted by peer.


Summary of changes from v2.4.27-pre4 to v2.4.27-pre5
============================================

<chrisg:etnus.com>:
  o x86-64 only: ia32entry.S reg changes dropped during debugging

<kevin.curtis:farsite.co.uk>:
  o [netdrvr wan] farsync update

<pelle:dsv.su.se>:
  o tulip PCI ID for D-Link DFE-680TXD

<vda:port.imtp.ilyichevsk.odessa.ua>:
  o fealnx #0: replace dev->base_addr with ioaddr; sync with 2.6
  o fealnx #1: replace magic constants with enums
  o fealnx #2: add 'static'; fix wrapped comment
  o fealnx #3: fix pointer substraction bug
  o fealnx #4: stop doing stop_nic_rx/writel(np->crvalue) in rest_rx_descriptors()
  o fealnx #5: introduce stop_nic_rxtx(), use it where makes sense
  o fealnx #6: Francois' fixes for low memory handling; remove free_one_rx_descriptor (not used anymore)
  o fealnx #7: Garzik fix (IIRC): add locking to tx_timeout
  o fealnx #8
  o fealnx #9
  o fealnx #10
  o fealnx #11

Andrew Morton:
  o 8139too not running s3 suspend/resume pci fix

Arjan van de Ven:
  o small change for scsi 2.6 header compatibility
  o [libata] Use standard headers from include/scsi, not drivers/scsi

Daniele Venzano:
  o [netdrvr sis900] fix ISA bridge detection
  o [netdrvr sis900] cosmetic header cleanups
  o [netdrvr sis900] fix missing netif_device_detach() in suspend

David S. Miller:
  o [TCP]: Fix build in 2.4.x with SCTP disabled
  o [PKT_SCHED]: Missing rta_len init in sch_delay
  o [SPARC64]: gcc-3.4.x build fixes

Don Fry:
  o 2.4.27-pre3 pcnet32 add static to two routines
  o 2.4.27-pre3 pcnet32 avoid hard hang with some chip variants
  o 2.4.27-pre3 pcnet32 correct 79C976 variant string
  o 2.4.27-pre3 pcnet32 fix boundary comparison bug
  o 2.4.27-pre3 pcnet32 remove timer and complexity
  o pcnet32: limit frames received during interrupt
  o pcnet32: fix bogus carrier errors with 79c973
  o pcnet32: correct printk for big-endian arch
  o pcnet32: avoid timeout with tcpdump
  o pcnet32: fix for patch 8 le16_to_cpu

François Romieu:
  o [netdrvr r8169] update to 2.6.x version; many fixes and changes

Ganesh Venkatesan:
  o e1000 1/7: Clear auto-mdix mode when forcing link to
  o e1000 2/7: Workaround for link LED staying ON even when
  o e1000 3/7: Determine Link Status correctly while using
  o e1000 4/7: Rewrite logic to estimate # of tx descriptors
  o e1000 6/7: ethtool_ops support
  o e1000 7/7: Error Logging support (enabled/disabled via
  o e100 1/1: Update to sync up version numbers

Jeff Garzik:
  o [sound i810] pci id cleanups
  o [libata] add new ->bmdma_setup hook
  o [libata] use new ->bmdma_{start,setup} method to properly support ATAPI
  o [libata] more ATAPI work - translate SCSI CDB to ATA PACKET
  o [libata] random minor bug fixes
  o [libata] kill ATA_QCFLAG_POLL flag
  o [libata] internal cleanups
  o [libata] minor stuff
  o [libata] handle non-data ATAPI commands via interrupt
  o [libata] DMADIR support
  o [libata] remove redundant use of ATA_QCFLAG_SG in ATAPI packet translation
  o [libata] SCSI->ATA simulator hacking: INQUIRY command
  o [libata] comments and constants
  o [libata] scsi simulator improvements: MODE SENSE, SEEK(6,10), REZERO_UNIT
  o [libata] replace ATA_QCFLAG_ATAPI with inline helper
  o [libata] polish DocBook docs a bit
  o [netdrvr sis900] sync with 2.6.x
  o [netdrvr e1000] use generic ethtool_ops provided in net/core/ethtool.c
  o [libata promise] revert broken taskfile delivery change
  o [libata scsi] ack SYNCHRONIZE CACHE command

Marcelo Tosatti:
  o Changed EXTRAVERSION to -pre5

Patrick McHardy:
  o [IPV4,6]: Fix off-by-one in max protocol-type check

Stefan Rompf:
  o [netdrvr b44] always restore PCI config on resume



Summary of changes from v2.4.27-pre3 to v2.4.27-pre4
============================================

<akepner:sgi.com>:
  o [TG3]: Fix ethtool -S
  o [TG3]: Make sure RX/TX flow control settings actually get set

<felixb:sgi.com>:
  o [XFS] Remove speculative preallocation from linvfs_get_block_core

<fsgqa:sgi.com>:
  o [XFS] Remove the 128K limitation on pagebuf_get_no_daddr() and allow the kmem_alloc() to fail.

<gnb:melbourne.sgi.com>:
  o [TG3]: Count rx_discards in rx_errors
  o [TG3]: Add more ethtool -S stats

<joshk:triplehelix.org>:
  o [SPARC64]: Pull in drivers/i2c/Config.in
  o [SPARC64]: Use $(CC) in check for egcs/gcc3
  o [SPARC64]: Backport some lvalue warning fixes from 2.6.x
  o [SPARC32]: Use $(LD) -V for NEW_GAS check

<jpk:sgi.com>:
  o [XFS] xfs_iomap_write_delay() was doing speculative allocations without checking if there were any real blocks already in the speculative allocation area. This could result in an allocation that overlaps pre-allocated space. This would result in an ASSERT failure in debug kernels, or invalid output from xfs_bmap.
  o [XFS] Add support for allocating additional file space in stripe width sized chunks. A new fstab/mount option, "swalloc" has been defined. If specified when mounting a striped file system, allocation requests will be rounded up to a stripe width if the file size is >= stripe width, and the data is being appended to eof. The 'swalloc' option is "off" by default.

<linux-kernel:vortech.net>:
  o [VLAN]: Use KERN_INFO for VLAN_INF

<mhuth:mvista.com>:
  o [IPV6]: Fix sock identity checking bug in tcp_ipv6_check_established

<michael.waychison:sun.com>:
  o cramfs use pagecache

<tony.cureington:hp.com>:
  o [TG3]: Add eeprom dump support

<zli4:cs.uiuc.edu>:
  o [SPARC]: Fix prom_prom_taken[].theres_more setting

Adrian Bunk:
  o [ATALK]: Fix modular build
  o [NET]: Missing MODULE_LICENSE in p8022 and psnap
  o SCSI ips compile error

Andi Kleen:
  o Fix pageattr cache flushing on P4 (thanks to Terence Ripperda)

Andrew Morton:
  o use-before-uninitialized value in ext3(2)_find_ goal

Christoph Hellwig:
  o [XFS] fix direct user memory dereference in bulkstat
  o [XFS] Use macros instead of inlines for spinlock wrappers to aid debugging.

Christophe Saout:
  o [CRYPTO]: Fix two scatterwalk problems

Dave Kleikamp:
  o JFS: error in __get_metapage caused by invalid size from ea_get
  o JFS: Don't return -EPERM for system xattrs
  o JFS: Make sizes of tid_t & lid_t consistent
  o JFS: [CHECKER] Memory leak on commonly executed path
  o JFS: [CHECKER] Dereference of NULL pointer if alloc_metapage fails
  o JFS: [CHECKER] if txCommit fails, don't call d_instantiate
  o JFS: fix hang in __get_metapage

David S. Miller:
  o [TG3]: Add 572x/575x PCI IDs
  o [TG3]: Add 5750 chip and PHY IDs
  o [TG3]: Prepare for 5750 support plus minor fixes
  o [TIGON3]: Detect and record PCI Express
  o [TG3]: PCI Express 5750_A0 chips need 5701_REG_WRITE_BUG treatment
  o [TG3]: Fix chiprev test in previous change
  o [TG3]: Do not set CLOCK_CTRL_DELAY_PCI_GRANT on PCI Express
  o [TG3]: Double delay after writing MAC_MI_MODE reg
  o [TG3]: Correct RDMAC/WDMAC mode settings on 5705/5750
  o [TG3]: Do not write stats coalescing ticks reg on 5705/5750
  o [TG3]: Update to 5788 capable 5705 TSO firmware, version 1.2.0
  o [TG3]: Update to non-5705 TSO firmware version 1.6.0
  o [TG3]: If asked to load TSO firmware on 5750, just return success
  o [TG3]: Add 5750 NVRAM programming plus 5704 MAC offset bug fix
  o [TG3]: Update LED programming to support 5750
  o [TG3]: Updated ASF handling for 5750
  o [TG3]: Include mss in every txd, not just the first, on 5750
  o [TG3]: On 5750 with TSO, need to set some special reg bits
  o [TG3]: Full chip reset tweaks for 5750
  o [TG3]: More 5750 chip reset tweaks
  o [TG3]: Do not enable slow clocks on 5750 with ASF
  o [TG3]: Rewrite dma_rwctrl settings to handle PCIX/PCIE
  o [TG3]: Add 572x/575x PCI IDs to driver table, update vers/reldate
  o [SPARC64]: Update defconfig
  o [TG3]: Update driver version and reldate
  o [NET]: Fix common typo, using NODEV when we mean ENODEV
  o [TG3]: Fix phantom spaces added to pci.ids file
  o [TCP]: Kill distance enforcement between tcp_mem[] elements
  o [TCP]: Abstract out all settings of tcp_opt->ca_state into a function
  o [TCP]: Backport Vegas support from 2.6.x
  o [TCP]: Backport BIC TCP from 2.6.x
  o [TCP]: Add tcp_default_win_scale sysctl
  o [TCP]: Add receiver side RTT estimation
  o [TCP]: Grow socket receive buffer based upon estimated sender window
  o [TCP]: More sysctl tweakings for rcvbuf stuff

Dean Roehrich:
  o [XFS] Dmapi preunmount event references null pointer

Dely Sy:
  o Documentation for new PCI hotplug drivers

Dmitry Torokhov:
  o [NET_SCHED]: Do not oops when user tries to attach a filter to a TBF qdisc

Eric Dean Moore:
  o MPT Fusion driver 2.05.16 update

Eric Sandeen:
  o [XFS] guard against unused var in new mutex_spinunlock #define

François Romieu:
  o CPUID Pentium-4E update and missing new line

Hideaki Yoshifuji:
  o [IPV6] ensure to evaluate the checksum for sockets with the IPV6_CHECKSUM option
  o [NET]: Prevent future missed updates of FOO_MAX macros

Hugh Dickins:
  o tmpfs surplus page miscounted

Jakub Bogusz:
  o missing include in drivers/sound/kahlua.c

Jeff Garzik:
  o [netdrvr tg3] netdev_priv
  o [netdrvr tg3] sync with 2.6.x

Len Brown:
  o [ACPI] delete IOAPIC-disable workaround on x86_64/VIA BTW. looks like 2.6 has an IOMMU disable workaround here that may be needed or VIA in 2.4.
  o [ACPI] revert button module unload fix (OSDL 2281) Cset exclude: len.brown@intel.com|ChangeSet|20040504154434|56458 Cset exclude: len.brown@intel.com|ChangeSet|20040428081912|57065 Cset exclude: len.brown@intel.com|ChangeSet|20040428054017|55837
  o [ACPI] remove /proc files before unloading modules from Sau Dan Lee, Zhenyu Wang http://bugzilla.kernel.org/show_bug.cgi?id=2705
  o [ACPI] x86_64 + ACPI + IOAPIC + PCI Interrupt Link -> IRQ 0 from Andy Currid

Marcelo Tosatti:
  o ext3_read_super: remove [un]lock_super  (Andrew Morton)
  o Changed EXTRAVERSION to -pre4

Nathan Scott:
  o Fix an incorrect email address in XFS maintainers section
  o [XFS] Remove unused transaction pointer from bulkstat
  o [XFS] Bump the kmalloc/vmalloc cutoff up to 128k
  o [XFS] Make uses of extended inode flags consistent, remove duplicated code
  o [XFS] Fix some compiler warnings, mark cmn_err as printflike
  o [XFS] Fixup a couple of incorrect xfs_trans_commit calls (bad flags/casts).
  o [XFS] Export/import tunable time intervals as centisecs not jiffies
  o [XFS] Switch all XFSDEBUG to DEBUG
  o [XFS] Fix a use-after-free during transaction commit when the log is in error state.
  o [XFS] Use set_current_state instead of direct current->state assignment
  o [XFS] Remove xfs_swappable code, its not useful on Linux
  o [XFS] Remove no-longer-used variable in log write code, and a dated comment.
  o [XFS] Remove unused xfs_trans_bhold_until_committed and related macros
  o [XFS] Rename a subdirectory to make life easier for people (esp

Oleg Drokin:
  o Fix possible memleaks in VIA IrDA driver

Patrick McHardy:
  o [IPV4]: Fix skb leak in igmpv3_newpack

Paul Mackerras:
  o Fix incorrect PT_FPSCR definition

Sridhar Samudrala:
  o [SCTP] Fix memset() parameter ordering
  o [SCTP] Fix accessing Gap Ack blocks array with a -ve index in sctp_outq_sack()
  o [SCTP]: Fix multihomed connection failures on 64-bit systems

Stephen Hemminger:
  o [BRIDGE]: Update bridge.txt
  o [TCP]: Add sysctl to turn off metrics caching
  o [TCP]: Add vegas sysctl docs

Timothy Shimmin:
  o [XFS] Change xfs_contig_bits to work on 32/64 and both endian styles

Trond Myklebust:
  o NFS client: Fix sillydelete()



Summary of changes from v2.4.27-pre2 to v2.4.27-pre3
============================================

<achurch:achurch.org>:
  o ext2fs sb= mount option fix

<chad.dupuis:hp.com>:
  o Fix HP HSG80 storage array entry on SCSI device list

<gtw:cs.bu.edu>:
  o [sound i810] silently ignore invalid PCM_ENABLE_xxx bits from userland

<jparmele:wildbear.com>:
  o ad1848: Fix crystal CS4235 mixer setup

<lkml:lievin.net>:
  o tipar char driver: wrong timeout value

<m.c.p:kernel.linux-systeme.com>:
  o copy WARN_ON() definition from 2.6

<ouellettes:videotron.ca>:
  o Remove extra semicolumn in arch/i386/mm/fault.c

<ross:datscreative.com.au>:
  o nVidia supplied nForce2 workaround

Andrew Morton:
  o sata_sx4.c warning fix

Arun Sharma:
  o ia64: make ia32 core dumps work

Bjorn Helgaas:
  o Cset exclude: arun.sharma@intel.com[helgaas]|ChangeSet|20040405163028|42090
  o ia64: Fix ia32 core dumps
  o ia64: Update defconfigs
  o ia64: Update defconfig to current generic config
  o ia64: Allow IO port space without EFI RT attribute
  o ia64: Define acpi_pci_disabled for recent ACPI update

Dave Kleikamp:
  o JFS: [CHECKER] Fix a possible null-pointer dereference
  o JFS: [CHECKER] Memory leak in jfs_link
  o JFS: [CHECKER] get rid of txAbortCommit
  o JFS: Avoid race invalidating metadata page
  o JFS: reduce stack usage
  o JFS: [CHECKER] More robust error recovery in add_index
  o JFS: module unload was not removing /proc/fs/jfs/

David Mosberger:
  o ia64: Fix typo in unwinder which could cause NULL-pointer dereferences

Don Fry:
  o pcnet32 whitespace only changes
  o pcnet32 support for 79C976
  o pcnet32 all printk under netif_msg
  o pcnet32 correct name display
  o pcnet32 add led blink capability
  o pcnet32 transmit performance fix
  o pcnet32 add register dump capability
  o pcnet32 timer to free tx skbs for 79C971/972

Gerd Knorr:
  o Fix V4L miss of range check oops

Herbert Xu:
  o [sound i810] fix wait queue race in drain_dac
  o [sound i810] fix race
  o [sound i810] remove bogus CIV_TO_LVI
  o [sound i810] clean up with macros
  o [sound i810] fix partial DMA transfers
  o [sound i810] fix playback SETTRIGGER
  o [sound i810] fix OSS fragments
  o [sound i810] remove divides on playback
  o [sound i810] fix drain_dac loop when signals_allowed==0
  o [sound i810] fix reads/writes % 4 != 0
  o [sound i810] fix deadlock in drain_dac

Jack Hammer:
  o ServeRAID driver update to 7.00.15: sync with v2.6

Jack Steiner:
  o ia64: fix HUGETLB null pointer dereference

Jeff Garzik:
  o [netdrvr b44] sync with 2.6.x version
  o Add dummy "__user" marker, for compat with 2.6.x
  o Rename get_current_user tmpvar to avoid namespace clash
  o [netdrvr] Add driver for IBM p/iSeries virtual ethernet adapters
  o [libata sata_sis] add new PCI id
  o [libata] Promise driver split part 1: clone to sx4
  o [libata] Promise driver split part 2: remove SX4 code from sata_promise
  o [libata] Promise driver split part 3: remove TX2/4 code from sata_sx4
  o [libata] Promise driver split part 4: common header
  o [libata] add ata_tf_{to,from}_fis helpers
  o [libata] clean up taskfile submission to hardware
  o [libata] remove call to WARN_ON(), 2.4 doesn't have this
  o [libata] move ATAPI startup from katad thread to workqueue thread
  o [libata] minor updates
  o [libata] move PIO data xfer from katad thread to workqueue thread
  o [libata] move probe execution from katad thread to workqueue thread
  o [libata] move ATAPI command initiation code from libata-scsi to libata-core
  o [libata] kill unnecessary include
  o [libata] internal cleanup: kill ata_pio_start
  o [libata] some work on the ATAPI path
  o [libata] Make sure to initialize PIO data xfer state
  o [libata] replace per-command semaphore with optional completion
  o [libata promise] make sure our schedule_timeout(N) are never with N==0
  o [libata] remove unused struct ata_engine
  o [libata sata_sx4] trivial: fix filename in header
  o [libata sata_sis] support SATA SCRs in PCI cfg space
  o [libata] preparation for writeback caching support
  o [libata] Maintainer annotations
  o [sound i810] sync with 2.6.x driver
  o [sound i810] bump driver to version 1.00

Jeremy Higdon:
  o sata_vsc initialization fix

Karol Kozimor:
  o acpi4asus 0.28 (Karol 'sziwan' Kozimor)

Krzysztof Halasa:
  o [netdrvr tulip] fix use-after-free

Len Brown:
  o [ACPI] delete unused CONFIG_ACPI_RELAXED_AML this code is included always -- or disable at boot w/ "acpi=strict
  o [ACPI] Delete IRQ2 "cascade" in ACPI IOAPIC mode no such concept exists in ACPI, frees IRQ2 for use.
  o [ACPI] enhance intr-src-override parsing to handle ES7000 http://bugme.osdl.org/show_bug.cgi?id=2520
  o [ACPI] allow use of IRQ2 in ACPI/IOAPIC mode http://bugzilla.kernel.org/show_bug.cgi?id=2564
  o [ACPI] No IRQ known... - using IRQ 255 (Bjarni Rúnar Einarsson) http://bugzilla.kernel.org/show_bug.cgi?id=2148
  o [ACPI] battery "charged" instead of "unknown" (Luming Yu) http://bugzilla.kernel.org/show_bug.cgi?id=1863
  o [ACPI] support button driver unload (Luming Yu) http://bugzilla.kernel.org/show_bug.cgi?id=2281
  o [ACPI] pci-link may not always be SHARED (SuSE via Luming Yu) http://bugzilla.kernel.org/show_bug.cgi?id=2404
  o [ACPI] pcibios_scan_root fix for IA64 (from IA64 tree via Luming Yu) http://bugzilla.kernel.org/show_bug.cgi?id=2130
  o [ACPI] toshiba_acpi driver if acpi_disabled (David Shaohua Li) http://bugzilla.kernel.org/show_bug.cgi?id=2465
  o [ACPI] button build fix
  o [ACPI] export symbols to button module
  o [ACPI] PCI Interrupt Link fixes
  o [ACPI] rmmod ACPI modules vs /proc from Anil S Keshavamurthy and David Shaohua Li http://bugzilla.kernel.org/show_bug.cgi?id=2457
  o [ACPI] enhance "pci=noacpi" to skip PCI probe (David Shaohua Li) add "acpi=noirq" to just disable IRQ config http://bugzilla.kernel.org/show_bug.cgi?id=1662
  o [ACPI] parse ACPI MCFG table and set pci_mmcfg_base_addr from 2.6 via Sundar/Dely Sy
  o [ACPI] handle _CRS outside _PRS -- even when non-zero avoid sharing IRQ12 http://bugzilla.kernel.org/show_bug.cgi?id=2665

Marcelo Tosatti:
  o Delete: drivers/char/amd7xx_tco.c
  o Matt Domsch: Lower printk severity for PCI devices with no PCI_CACHE_LINE_SIZE implemented
  o Changed EXTRAVERSION to -pre3

Mike Miller:
  o cciss update: support for two new controllers
  o Fix cciss bug in proc reporting

Oleg Drokin:
  o [2.4] Make reiserfs not to crash on oom during mount

Pavel Machek:
  o Cleanups for b44

Pete Zaitcev:
  o tiglusb bug fixes (usb_clear_halt, usb_sndbulkpipe)
  o USB: Check results of copy_to_user in hiddev
  o USB: Fix memory leaks in speedtch
  o USB: Fix IBM USB memory key unusual dev entry

Richard Curnow:
  o Patch [sh64]: Fix liveness dependency constraints in _syscalln()
  o Fix Cayman PCI IRQ routing for PCI cards containing bridges
  o Patch [sh64]: Fix approach for calibrating CPU MHz so it works even if the I-cache is disabled
  o Patch [sh64]: Clean-up handling of the SR.WATCH bit
  o Patch [sh64]: Remove obsolete CONFIG_UNCACHED_MEMORY_OFFSET option
  o Patch [sh64]: clean-up to allow building on Cayman board without configuring PCI support
  o Patch [sh64]: Handle base of RAM not being 512Mb aligned
  o Patch [sh64]: Allow greater choice over the effective address at which the kernel runs

Sridhar Samudrala:
  o [SCTP] Fix bugs in handling overlapping INIT and peer restart over a multihomed association.
  o [SCTP] Rename SCTP_ADDR_REACHABLE as SCTP_ADDR_AVAILABLE to be consistent with the SCTP sockets API draft.

Stephen Hemminger:
  o SCTP crc table can be static const

Stéphane Eranian:
  o ia64: switch /proc/perfmon to seq_file avoid buffer overflows

Zwane Mwaikambo:
  o remove amd7(saucy)_tco



Summary of changes from v2.4.27-pre1 to v2.4.27-pre2
============================================

<aj:andaco.de>:
  o [TG3]: Fix typo in TG3_TSO_FW_RODATA_ADDR definition

<gandalf:winds.org>:
  o Blue line in nVidia framebuffer (rivafb)

<jkmaline:cc.hut.fi>:
  o [CRYPTO]: Add suuport for keyed digests
  o [CRYPTO]: Add Michael MIC algorithm

<jpk:sgi.com>:
  o [XFS] Correct the (file size >= stripe unit) check inside xfs_iomap_write_delay.  It was comparing the file size, in bytes, against the stripe unit size, in FSBs.

<mchan:broadcom.com>:
  o [TG3]: Jumbo frames and FTQ reset patch
  o [TG3]: Fix jumbo frame PHY programming

Andrea Arcangeli:
  o Fix page-free-at-IRQ context BUG

Bartlomiej Zolnierkiewicz:
  o IDE PCI: fix support for PIO modes w/o IORDY flow control
  o serverworks.c: fix DMA for OSB4 (Patrick Wildi)
  o generic PCI IDE support for more Toshiba Piccolo chips (Daniel Drake)

Chas Williams:
  o [ATM]: get atm_guess_pdu2truesize() right
  o [ATM]: [nicstar] using dev_alloc_skb() (reported by Johnston,
  o [ATM]: [fore200e] 0.3e version by Christophe Lizzi (lizzi@cnam.fr)
  o [ATM]: [fore200e] make tasklet use configurable

Chris Wright:
  o [IPV4]: Fix return value on MCAST_MSFILTER error case

Christoph Hellwig:
  o [XFS] use kmem_alloc for noaddr buffers
  o [XFS] kill the pagebuf vs xfs_buf confusion
  o [XFS] really kill the pagebuf vs xfs_buf confusion
  o [XFS] clarify pagebuf page lookup logic
  o [XFS] cleanup pagebuf flag usage and simplify pagebuf_free
  o [XFS] close external blockdevice after final flush

David S. Miller:
  o [TG3]: Just completely delete the disabled FTQ reset code
  o [SPARC64]: Kill cast as lvalue usage in pgd/pmd macros
  o [SPARC64]: Fix rest of cast as lvalue usage in sparc64 code
  o [TG3]: Kill 'force' arg to tg3_phy_reset, it is always set
  o [TG3]: At start of tg3_phy_copper_begin, force phy out of loopback mode
  o [TG3]: Do not allow illegal ethtool advertisement bits
  o [SPARC64]: Export prom_palette
  o [TG3]: Add missing 5704 BX workaround, and fix typo in autoneg fix
  o [TG3]: Set GRC_MISC_CFG prescaler more safely
  o [TG3]: Fix serdes cfg programming on 5704
  o [TG3]: When link is down, set stats coalescing ticks to zero
  o [TG3]: Wait a bit for BMSR/BMCR bits to settle in PHY setup
  o [TG3]: Verify link advertisement correctly on 10/100 only chips
  o [TG3]: All 5705 chips need PHY reset on link-down
  o [TG3]: More PHY programming fixes
  o [TG3]: Bump driver version and reldate
  o [TG3]: Print list of important probed capabilities at driver load
  o [TG3]: Two PHY fixes
  o [TG3]: Kill uninitialized var warning
  o [TG3]: Reset fixes
  o [TG3]: Update driver version and release date
  o [SPARC]: Use 64 for KERN_SPARC_SCONS_PWROFF to prevent 2.6.x conflicts
  o [TG3]: Update driver version and reldate
  o [TG3]: Undo comment typo fix, it was wrong
  o [SPARC64]: Fix zero-extension issues wrt. {pgd,pmd}_val()
  o [SPARC64]: Update defconfig

David Stevens:
  o [IPV4]: Fix IGMP version number and timer printing for procfs

Dely Sy:
  o SHPC and PCI Express hot-plug drivers for 2.4 kernel
  o PCI Hotplug: SHPC & PCI-E hot-plug fixes

Eric Brower:
  o [SPARC]: Add sysctl to control serial console power-off restrictions
  o [SPARC64]: HDIO_DRIVE_TASK is a compatible ioctl

Eric Sandeen:
  o [XFS] Use pgoff_t for page indices, and remove some other type confusion
  o [XFS] New PFLAGS_RESTORE_FSTRANS macro to restore only FSTRANS state from saved state.

Geert Uytterhoeven:
  o [NET]: Make pktgen depend upon procfs
  o NCR53C9x unused SCp.have_data_in
  o M68k TLB fixes
  o Amiga A2065 Ethernet debug

Grant Grundler:
  o [TG3]: Fix comment typo

Harald Welte:
  o [NETFILTER]: Add more debug info to TFTP helper

Hideaki Yoshifuji:
  o [IPV6]: Mark MLDv2 report as known
  o [IPV6]: Use IANA icmpv6 type for MLDv2 report

Jakub Bogusz:
  o rivafb 16bpp text background colour fix

Jamal Hadi Salim:
  o [NET_SCHED]: Check for NULL opt in dsmark_init

Jeff Garzik:
  o [TG3]: Dump NIC-specific statistics via ethtool

Jon Oberheide:
  o [CRYPTO]: ARC4 config help clarification

Marcel Holtmann:
  o [Bluetooth] Allow normal users to release the previous created TTY
  o [Bluetooth] Fix race in RX complete routine of the USB drivers
  o [Bluetooth] Make use of request_firmware() for the 3Com driver
  o [Bluetooth] Add UART protocol id's for 3-Wire and H4DS

Marcelo Tosatti:
  o Andrew Morton: __free_pages_ok() stress testing
  o Changed EXTRAVERSION to -pre2

Matt Porter:
  o ppc32: fix head_44x.S copyrights

Nathan Scott:
  o [XFS] Fix a very hard-to-hit, small-block-size only corruption
  o [XFS] Fix delayed write buffer handling to use the correct list interfaces, add validity checks, remove unused code, fix comments.
  o [XFS] Make buffer error checking consistent, add a value range check
  o [XFS] Return the right error code on an ACL xattr version mismatch
  o [XFS] Allow xfsbufd flush intervals to take immediate effect after changing the flush sysctl value.  Fix from Bart Samwel.
  o [XFS] Clear the superblock dirty flag after flushing the log in sync_super.
  o [XFS] Fix vmtruncate abuse in the XFS setattr ATTR_SIZE operation
  o [XFS] make return value type for read() really ssize_t
  o [XFS] Define a new superblock field for more feature bits
  o [XFS] Fix debug builds - need sb_features2 details in the endian translation code.
  o Remove a bk ignored XFS cvs directory, accidentally added
  o Fix typo in delayed allocate buffer count reporting (sysrq-m)
  o Minor updates to XFS documentation
  o [XFS] Fix fsync regression resulting from moving data flushing out from under the IOLOCK.
  o [XFS] Remove extraneous vmtruncate call, missed in earlier merge
  o [XFS] Remove xfs_iaccess checks on security extended attribute namespace, done outside the filesystem.

Russell Cattelan:
  o [XFS] Fix for the xfs dir2 rebalance bug

Rusty Russell:
  o [NETFILTER]: Missing ip_rt_put in ipt_MASQUERADE

Sridhar Samudrala:
  o [SCTP] Fix typo in entry name of the remove_proc_entry() call
  o [SCTP] Update sctp_ulpevent structure to include assoc pointer and
  o [SCTP] Avoid the use of constant SCTP_IP_OVERHEAD to determin the max data size in a SCTP packet.
  o [SCTP] Cleanup sctp_packet and sctp_outq infrastructure
  o [SCTP] Partial Reliability Extension support
  o [SCTP] Propagate error from sctp_proc_init. (Olaf Kirch)

Stephen C. Tweedie:
  o fix O(N^2) dquot sync behaviour

Takayoshi Kochi:
  o PCI Hotplug: acpiphp unable to power off slots

Takayoshi Kouchi:
  o PCI Hotplug: acpiphp cleanup patch for 2.4.23-pre4

William Lee Irwin III:
  o Joel Becker: Fix summit crash: cpu_present_to_apicid() bound checking

Zwane Mwaikambo:
  o fix module load with gcc3.3.3
  o Fix typo in include/linux/compiler.h


Summary of changes from v2.4.26 to v2.4.27-pre1
============================================

<khawar.chaudhry:amd.com>:
  o Update amd8111 net driver

<kirillx:7ka.mipt.ru>:
  o Fix potential memory leak in devpts
  o Fix potential memory access to free memory in /proc handling

<lkml:lievin.net>:
  o tipar char driver (divide by zero)

<maragato:gmx.net>:
  o Add ATI IGP 345M rev2 ID's

<mhf:linuxmail.org>:
  o Update codingstyle to 2.6 level

<sezero:superonline.com>:
  o megaraid2 compilation fix

<tmattox:engr.uky.edu>:
  o [netdrvr tulip] add MII support for Comet chips

<vda:port.imtp.ilyichevsk.odessa.ua>:
  o gcc3 does not inline some functions

Atul Mukker:
  o megaraid2 driver version 2.10.3

Chris Wright:
  o e1000: fix probable security hole

Don Fry:
  o resync pcnet32.c with 2.6.x
  o netdevice.h add netif_msg_init helper
  o pcnet32 fix hang/crash with loopback test

Ganesh Venkatesan:
  o e100: NFS/TCO related Firmware update
  o e100: change log + version update
  o e100: use new API, SET_NETDEV_DEV, rx_bytes stat to include MAC header fix
  o e1000: ethtool set/get ring param support
  o e1000: backoff Tanacross missed interrupt workaround
  o e1000: Changed E1000_COLLISION_THRESHOLD from 16 to 15
  o e1000: use E1000_PBA_BYTES_SHIFT instead of E1000_TX_FIFO_SIZE
  o e1000: remove polarity reversal workaround for forced 10H/10F links
  o e1000: fix eeprom update to include e1000_standby_eeprom
  o e1000: new bit definitions, fix comments
  o e1000: ethtool set/get eeprom fixes
  o e1000: use new API, SET_NETDEV_DEV, check register_netdev retval
  o e1000: all other white space fixes, changelog
  o e1000: Disable TSO - till TSO related Tx hangs are root caused/fixed
  o e1000: msec_delay fix
  o e1000: phy fix, and cleanup

Jeff Garzik:
  o [netdrvr 8139cp] trivial syncing with 2.6.x
  o [netdrvr 8139cp] rearrange priv struct, add cacheline-align markers
  o [netdrvr 8139cp] locking cleanups
  o [NET] forward-compat definition of netdev_priv()
  o [netdrvr 8139cp] minor cleanups
  o [netdrvr 8139cp] use netdev_priv()
  o [netdrvr 8139cp] complete 64-bit DMA (PCI DAC) support
  o [netdrvr 8139cp] better dev->close() handling, and misc related stuff
  o [netdrvr natsemi] correct DP83816 IntrHoldoff register offset
  o [netdrvr tulip] remove ChangeLog file, we have BitKeeper logs now
  o Add SATA support
  o Add Promise SX8 SATA driver
  o [netdrvr via-rhine] Fix MII phy scanning bug, whitespace cleanups

Marcelo Tosatti:
  o Herbert Xu: Delete unused drivers/sound/Hwmcode.h, drivers/sound/724hwmcode.h
  o Cset exclude: loftin@ldl.fc.hp.com|ChangeSet|20040414205510|54931
  o Changed EXTRAVERSION to 2.4.27-pre1
  o Fix drmP.h fix typo

Meelis Roos:
  o Kaupo Arulo: only use set_max when it is present

Paul Gortmaker:
  o [netdrvr 8390] Fix 8390 log spam

Pavel Roskin:
  o Tulip endianess fix

Pete Zaitcev:
  o Improve USB printer locking
  o More USB storage locking fixes

Randy Dunlap:
  o drmP.h doesn't need local cmpxchg() and __cmpxchg()

Scott Feldman:
  o Update MAINTAINERS with new e100/e100 maintainers

Zwane Mwaikambo:
  o fix module load with gcc3.3.3

eth*: transmit timed out since .27 (was: linux-2.4.27 released)

[Oliver Feiler]

[-- Attachment #1.1: body text --]
[-- Type: text/plain, Size: 2513 bytes --]

Hi,

I've upgraded a server from .26 to .27, but ran into problems with the network 
cards.

The kernel throws a lot of errors into the syslog and the net devices don't 
work:
Aug 10 13:39:25 spot kernel: NETDEV WATCHDOG: eth0: transmit timed out
Aug 10 13:39:26 spot kernel: NETDEV WATCHDOG: eth1: transmit timed out
Aug 10 13:39:26 spot kernel: eth1: Transmit timeout, status 00000004 00000249 
Aug 10 13:39:34 spot kernel: NETDEV WATCHDOG: eth1: transmit timed out
Aug 10 13:39:34 spot kernel: eth1: Transmit timeout, status 00000004 00000241 
Aug 10 13:39:42 spot kernel: NETDEV WATCHDOG: eth1: transmit timed out
Aug 10 13:39:42 spot kernel: eth1: Transmit timeout, status 00000004 00000240 
[...]

and:
Aug 10 13:39:25 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3, 
ISR=0x3, t=515.
Aug 10 13:40:25 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3, 
ISR=0x3, t=5015.
Aug 10 13:40:40 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3, 
ISR=0x3, t=1014.
[...]

The system has three network cards.
eth0: SIS900 (sis900.c)
eth1: RTL-8029 (ne2k-pci.c)
eth2: onboard VIA VT6102 Rhine-II (via-rhine.c)

eth0 and eth1 share the same interrupt

           CPU0       
  0:      91986          XT-PIC  timer
  1:        935          XT-PIC  keyboard
  2:          0          XT-PIC  cascade
  8:          1          XT-PIC  rtc
  9:          0          XT-PIC  acpi
 10:      25109          XT-PIC  via82cxxx, usb-uhci, usb-uhci, eth0, eth1
 11:         24          XT-PIC  usb-uhci, eth2
 14:       7523          XT-PIC  ide0
 15:       7021          XT-PIC  ide1
NMI:          0 
ERR:          0

That was not a problem in .26 however. Though it seems to be the cause of the 
problem (lost interrupt)? The hardware this is all running on is an Asrock 
K7VM4 mainboard. The system is booted with "pci=noacpi" (ACPI, no APM). 
Otherwise IRQ255 is assigned to IDE and someone told me the noacpi parameter 
would fix the board's braindead BIOS.

Either way .27 doesn't want to boot. I've attached dmesg from a running 2.4.26 
kernel and the config used for 2.4.27.

Other postings I've found say that the transmit timeouts mean that the 
lowlevel ethernet connection between the NICs broke. But this works fine in 
earlier kernels and only eth0 and eth1 which share an interrupt are affected. 
I'd be glad for any more suggestions on what might be causing this. :)

Thanks,
	Oliver


-- 
Oliver Feiler  -  http://kiza.kcore.de/

[-- Attachment #1.2: dmesg --]
[-- Type: text/plain, Size: 9313 bytes --]

Linux version 2.4.26 (root@spot) (gcc version 3.3.4) #3 Mon Jul 5 15:32:52 CEST 2004
BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
 BIOS-e820: 000000000009fc00 - 00000000000a0000 (reserved)
 BIOS-e820: 00000000000d0000 - 00000000000d6000 (reserved)
 BIOS-e820: 00000000000f0000 - 0000000000100000 (reserved)
 BIOS-e820: 0000000000100000 - 000000000f7f0000 (usable)
 BIOS-e820: 000000000f7f0000 - 000000000f7f8000 (ACPI data)
 BIOS-e820: 000000000f7f8000 - 000000000f800000 (ACPI NVS)
 BIOS-e820: 00000000fec00000 - 00000000fec01000 (reserved)
 BIOS-e820: 00000000fee00000 - 00000000fee01000 (reserved)
 BIOS-e820: 00000000fff80000 - 0000000100000000 (reserved)
247MB LOWMEM available.
On node 0 totalpages: 63472
zone(0): 4096 pages.
zone(1): 59376 pages.
zone(2): 0 pages.
ACPI: RSDP (v000 AMI                                       ) @ 0x000fa620
ACPI: RSDT (v001 AMIINT VIA_K7   0x00000010 MSFT 0x00000097) @ 0x0f7f0000
ACPI: FADT (v001 AMIINT VIA_K7   0x00000011 MSFT 0x00000097) @ 0x0f7f0030
ACPI: MADT (v001 AMIINT VIA_K7   0x00000009 MSFT 0x00000097) @ 0x0f7f00c0
ACPI: DSDT (v001    VIA    K7VT4 0x00001000 MSFT 0x0100000d) @ 0x00000000
Kernel command line: BOOT_IMAGE=Linux.old ro root=900 pci=noacpi
Initializing CPU#0
Detected 599.436 MHz processor.
Console: colour VGA+ 80x25
Calibrating delay loop... 1196.03 BogoMIPS
Memory: 248184k/253888k available (1668k kernel code, 5316k reserved, 578k data, 92k init, 0k highmem)
Dentry cache hash table entries: 32768 (order: 6, 262144 bytes)
Inode cache hash table entries: 16384 (order: 5, 131072 bytes)
Mount cache hash table entries: 512 (order: 0, 4096 bytes)
Buffer cache hash table entries: 16384 (order: 4, 65536 bytes)
Page-cache hash table entries: 65536 (order: 6, 262144 bytes)
CPU: L1 I Cache: 64K (64 bytes/line), D cache 64K (64 bytes/line)
CPU: L2 Cache: 64K (64 bytes/line)
Intel machine check architecture supported.
Intel machine check reporting enabled on CPU#0.
CPU:     After generic, caps: 0183fbff c1c7fbff 00000000 00000000
CPU:             Common caps: 0183fbff c1c7fbff 00000000 00000000
CPU: AMD Duron(tm)  stepping 00
Enabling fast FPU save and restore... done.
Checking 'hlt' instruction... OK.
POSIX conformance testing by UNIFIX
mtrr: v1.40 (20010327) Richard Gooch (rgooch@atnf.csiro.au)
mtrr: detected mtrr type: Intel
ACPI: Subsystem revision 20040326
PCI: PCI BIOS revision 2.10 entry at 0xfdae1, last bus=1
PCI: Using configuration type 1
ACPI: IRQ9 SCI: Edge set to Level Trigger.
ACPI: Interpreter enabled
ACPI: Using PIC for interrupt routing
ACPI: System [ACPI] (supports S0 S1 S4 S5)
ACPI: PCI Root Bridge [PCI0] (00:00)
PCI: Probing PCI hardware (bus 00)
ACPI: PCI Interrupt Routing Table [\_SB_.PCI0._PRT]
ACPI: Power Resource [URP1] (off)
ACPI: Power Resource [URP2] (off)
ACPI: Power Resource [FDDP] (off)
ACPI: Power Resource [LPTP] (off)
ACPI: PCI Interrupt Link [LNKA] (IRQs 3 4 5 6 7 10 *11 12 14 15)
ACPI: PCI Interrupt Link [LNKB] (IRQs 3 4 5 6 7 *10 11 12 14 15)
ACPI: PCI Interrupt Link [LNKC] (IRQs 3 4 5 6 7 *10 11 12 14 15)
ACPI: PCI Interrupt Link [LNKD] (IRQs 3 4 5 6 7 *10 11 12 14 15)
PCI: Probing PCI hardware
PCI: Using IRQ router default [1106/3177] at 00:11.0
PCI: Hardcoded IRQ 14 for device 00:11.1
Linux NET4.0 for Linux 2.4
Based upon Swansea University Computer Society NET3.039
Initializing RT netlink socket
Starting kswapd
Journalled Block Device driver loaded
Installing knfsd (copyright (C) 1996 okir@monad.swb.de).
ACPI: Power Button (FF) [PWRF]
ACPI: Sleep Button (CM) [SLPB]
ACPI: Processor [CPU1] (supports C1)
pty: 256 Unix98 ptys configured
Serial driver version 5.05c (2001-07-08) with MANY_PORTS SHARE_IRQ SERIAL_PCI enabled
ttyS00 at 0x03f8 (irq = 4) is a 16550A
ttyS01 at 0x02f8 (irq = 3) is a 16550A
Real Time Clock Driver v1.10f
FDC 0 is a post-1991 82077
ne2k-pci.c:v1.02 10/19/2000 D. Becker/P. Gortmaker
  http://www.scyld.com/network/ne2k-pci.html
eth0: RealTek RTL-8029 found at 0xe800, IRQ 10, 00:00:E8:5C:2D:AA.
sis900.c: v1.08.06 9/24/2002
eth1: SiS 900 Internal MII PHY transceiver found at address 1.
eth1: Using transceiver found at address 1 as default
eth1: SiS 900 PCI Fast Ethernet at 0xec00, IRQ 10, 00:c0:ca:16:4c:b6.
PPP generic driver version 2.4.2
PPP Deflate Compression module registered
PPP BSD Compression module registered
Uniform Multi-Platform E-IDE driver Revision: 7.00beta4-2.4
ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
VP_IDE: IDE controller at PCI slot 00:11.1
PCI: Hardcoded IRQ 14 for device 00:11.1
VP_IDE: chipset revision 6
VP_IDE: not 100% native mode: will probe irqs later
ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
VP_IDE: VIA vt8235 (rev 00) IDE UDMA133 controller on pci00:11.1
    ide0: BM-DMA at 0xfc00-0xfc07, BIOS settings: hda:DMA, hdb:pio
    ide1: BM-DMA at 0xfc08-0xfc0f, BIOS settings: hdc:DMA, hdd:pio
hda: WDC WD800BB-00CAA1, ATA DISK drive
blk: queue c0371b40, I/O limit 4095Mb (mask 0xffffffff)
hdc: ST380011A, ATA DISK drive
blk: queue c0371f94, I/O limit 4095Mb (mask 0xffffffff)
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
ide1 at 0x170-0x177,0x376 on irq 15
hda: attached ide-disk driver.
hda: host protected area => 1
hda: 156301488 sectors (80026 MB) w/2048KiB Cache, CHS=9729/255/63, UDMA(100)
hdc: attached ide-disk driver.
hdc: host protected area => 1
hdc: 156301488 sectors (80026 MB) w/2048KiB Cache, CHS=9729/255/63, UDMA(100)
Partition check:
 hda: hda1 hda2 hda3
 hdc: hdc1 hdc2 hdc3
Via 686a/8233/8235 audio driver 1.9.1-ac3
via82cxxx: Six channel audio available
PCI: Setting latency timer of device 00:11.5 to 64
ac97_codec: AC97  codec, id: CMI97 (CMedia)
AC97 codec does not have proper volume support.
via82cxxx: Codec rate locked at 48Khz
via82cxxx: board #1 at 0xD800, IRQ 10
usb.c: registered new driver hub
host/usb-uhci.c: $Revision: 1.275 $ time 15:33:03 Jul  5 2004
host/usb-uhci.c: High bandwidth mode enabled
host/usb-uhci.c: USB UHCI at I/O 0xe400, IRQ 10
host/usb-uhci.c: Detected 2 ports
usb.c: new USB bus registered, assigned bus number 1
hub.c: USB hub found
hub.c: 2 ports detected
host/usb-uhci.c: USB UHCI at I/O 0xe000, IRQ 10
host/usb-uhci.c: Detected 2 ports
usb.c: new USB bus registered, assigned bus number 2
hub.c: USB hub found
hub.c: 2 ports detected
host/usb-uhci.c: USB UHCI at I/O 0xdc00, IRQ 11
host/usb-uhci.c: Detected 2 ports
usb.c: new USB bus registered, assigned bus number 3
hub.c: USB hub found
hub.c: 2 ports detected
host/usb-uhci.c: v1.275:USB Universal Host Controller Interface driver
i2c-core.o: i2c core module version 2.8.3 (20040115)
i2c-dev.o: i2c /dev entries driver module version 2.8.3 (20040115)
i2c-proc.o version 2.8.3 (20040115)
md: raid1 personality registered as nr 3
md: md driver 0.90.0 MAX_MD_DEVS=256, MD_SB_DISKS=27
md: Autodetecting RAID arrays.
 [events: 000005f4]
 [events: 000005f4]
md: autorun ...
md: considering hdc3 ...
md:  adding hdc3 ...
md:  adding hda3 ...
md: created md0
md: bind<hda3,1>
md: bind<hdc3,2>
md: running: <hdc3><hda3>
md: hdc3's event counter: 000005f4
md: hda3's event counter: 000005f4
md: RAID level 1 does not need chunksize! Continuing anyway.
md0: max total readahead window set to 124k
md0: 1 data-disks, max readahead per data-disk: 124k
raid1: device hdc3 operational as mirror 1
raid1: device hda3 operational as mirror 0
raid1: raid set md0 active with 2 out of 2 mirrors
md: updating md0 RAID superblock on device
md: hdc3 [events: 000005f5]<6>(write) hdc3's sb offset: 77858944
md: hda3 [events: 000005f5]<6>(write) hda3's sb offset: 77851776
md: ... autorun DONE.
Initializing Cryptographic API
NET4: Linux TCP/IP 1.0 for NET4.0
IP Protocols: ICMP, UDP, TCP, IGMP
IP: routing cache hash table of 2048 buckets, 16Kbytes
TCP: Hash tables configured (established 16384 bind 32768)
ip_conntrack version 2.1 (1983 buckets, 15864 max) - 288 bytes per conntrack
ip_tables: (C) 2000-2002 Netfilter core team
NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
kjournald starting.  Commit interval 5 seconds
EXT3-fs: mounted filesystem with ordered data mode.
VFS: Mounted root (ext3 filesystem) readonly.
Freeing unused kernel memory: 92k freed
Adding Swap: 240964k swap-space (priority -1)
Adding Swap: 249976k swap-space (priority -2)
EXT3 FS 2.4-0.9.19, 19 August 2002 on md(9,0), internal journal
i2c-viapro.o version 2.8.3 (20040115)
i2c-dev.o: Registered 'SMBus Via Pro adapter at 0400' as minor 0
i2c-isa.o version 2.8.3 (20040115)
i2c-dev.o: Registered 'ISA main adapter' as minor 1
w83627hf.o version 2.8.3 (20040115)
via-rhine.c:v1.10-LK1.1.19  July-12-2003  Written by Donald Becker
  http://www.scyld.com/network/via-rhine.html
eth2: VIA VT6102 Rhine-II at 0xd400, 00:0b:6a:2b:48:84, IRQ 11.
eth2: MII PHY found at address 1, status 0x786d advertising 05e1 Link 45e1.
kjournald starting.  Commit interval 5 seconds
EXT3 FS 2.4-0.9.19, 19 August 2002 on ide0(3,1), internal journal
EXT3-fs: mounted filesystem with ordered data mode.
eth2: Setting full-duplex based on MII #1 link partner capability of 45e1.
eth1: Media Link On 100mbps half-duplex 
HTB init, kernel part version 3.16

[-- Attachment #1.3: config-2.4.27.gz --]
[-- Type: application/x-gzip, Size: 5287 bytes --]

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: linux-2.4.27 released

[William Lee Irwin III]

On Sat, Aug 07, 2004 at 04:28:27PM -0700, Marcelo Tosatti wrote:
> - 2.4.27-rc6 was released as 2.4.27 with no changes.
> Here is a list of the most important security issues fixed by this release:
> CAN-2004-0495 (Al Viro sparse fixes)
> CAN-2004-0497 (users could modify group ID of arbitrary files on the system)
> CAN-2004-0535 (e1000 minor info leak)
> CAN-2004-0685 (backported Conectiva usb sparse fixes)
> CAN-2004-0415 (file offset pointer handling race)
> CAN-2004-0565 (information leak ia64)

This patch by nature corrects two apparent bugs which are really one
bug. p->mm can become NULL while traversing the tasklist. The two
effects are first that kernel threads appear to be killed. The second
is that the OOM killing process fails to actually shoot down all threads
of the chosen process, and so fails to reclaim the memory it intended to.
oom_kill_task() consists primarily of the expansion of the 2.6 inline
function get_task_mm().

Index: linux-2.4/mm/oom_kill.c
===================================================================
--- linux-2.4.orig/mm/oom_kill.c	2004-06-23 19:30:21.000000000 -0700
+++ linux-2.4/mm/oom_kill.c	2004-06-23 19:52:25.000000000 -0700
@@ -141,7 +141,7 @@
  * CAP_SYS_RAW_IO set, send SIGTERM instead (but it's unlikely that
  * we select a process with CAP_SYS_RAW_IO set).
  */
-void oom_kill_task(struct task_struct *p)
+static void __oom_kill_task(struct task_struct *p)
 {
 	printk(KERN_ERR "Out of Memory: Killed process %d (%s).\n", p->pid, p->comm);
 
@@ -161,6 +161,26 @@
 	}
 }
 
+static struct mm_struct *oom_kill_task(struct task_struct *p)
+{
+	struct mm_struct *mm;
+
+	task_lock(p);
+	mm = p->mm;
+	if (mm) {
+		spin_lock(&mmlist_lock);
+		if (atomic_read(&mm->mm_users))
+			atomic_inc(&mm->mm_users);
+		else
+			mm = NULL;
+		spin_unlock(&mmlist_lock);
+	}
+	task_unlock(p);
+	if (mm)
+		__oom_kill_task(p);
+	return mm;
+}
+
 /**
  * oom_kill - kill the "best" process when we run out of memory
  *
@@ -172,21 +192,27 @@
 static void oom_kill(void)
 {
 	struct task_struct *p, *q;
+	struct mm_struct *mm;
 
+retry:
 	read_lock(&tasklist_lock);
 	p = select_bad_process();
 
 	/* Found nothing?!?! Either we hang forever, or we panic. */
 	if (p == NULL)
 		panic("Out of memory and no killable processes...\n");
-
+	mm = oom_kill_task(p);
+	if (!mm) {
+		read_unlock(&tasklist_lock);
+		goto retry;
+	}
 	/* kill all processes that share the ->mm (i.e. all threads) */
 	for_each_task(q) {
-		if (q->mm == p->mm)
-			oom_kill_task(q);
+		if (q->mm == mm)
+			__oom_kill_task(q);
 	}
 	read_unlock(&tasklist_lock);
-
+	mmput(mm);
 	/*
 	 * Make kswapd go out of the way, so "p" has a good chance of
 	 * killing itself before someone else gets the chance to ask

Re: linux-2.4.27 released

[William Lee Irwin III]

On Thu, Aug 12, 2004 at 11:17:12AM -0700, William Lee Irwin III wrote:
> This patch by nature corrects two apparent bugs which are really one
> bug. p->mm can become NULL while traversing the tasklist. The two
> effects are first that kernel threads appear to be killed. The second
> is that the OOM killing process fails to actually shoot down all threads
> of the chosen process, and so fails to reclaim the memory it intended to.
> oom_kill_task() consists primarily of the expansion of the 2.6 inline
> function get_task_mm().

Incremental atop the mm reference patch:

out_of_memory() attempts to determine whether one jiffies-valued
variable refers to a point in time preceding another jiffies-valued
variable, but does not do so in a jiffies wrap -safe fashion. The
following patch corrects this by using the expansion of the 2.6
macro time_after() to check this condition.

Index: linux-2.4/mm/oom_kill.c
===================================================================
--- linux-2.4.orig/mm/oom_kill.c	2004-06-23 19:41:08.000000000 -0700
+++ linux-2.4/mm/oom_kill.c	2004-06-23 19:50:59.000000000 -0700
@@ -289,7 +289,7 @@
 	spin_lock(&oom_lock);
 
 reset:
-	if (first < now)
+	if ((long)first - (long)now < 0)
 		first = now;
 	count = 0;
 

Re: eth*: transmit timed out since .27 (was: linux-2.4.27 released)

[Marcelo Tosatti]

Hi Oliver,

On Tue, Aug 10, 2004 at 02:23:34PM +0200, Oliver Feiler wrote:
> Hi,
> 
> I've upgraded a server from .26 to .27, but ran into problems with the network 
> cards.
> 
> The kernel throws a lot of errors into the syslog and the net devices don't 
> work:
> Aug 10 13:39:25 spot kernel: NETDEV WATCHDOG: eth0: transmit timed out
> Aug 10 13:39:26 spot kernel: NETDEV WATCHDOG: eth1: transmit timed out
> Aug 10 13:39:26 spot kernel: eth1: Transmit timeout, status 00000004 00000249 
> Aug 10 13:39:34 spot kernel: NETDEV WATCHDOG: eth1: transmit timed out
> Aug 10 13:39:34 spot kernel: eth1: Transmit timeout, status 00000004 00000241 
> Aug 10 13:39:42 spot kernel: NETDEV WATCHDOG: eth1: transmit timed out
> Aug 10 13:39:42 spot kernel: eth1: Transmit timeout, status 00000004 00000240 
> [...]
> 
> and:
> Aug 10 13:39:25 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3, 
> ISR=0x3, t=515.
> Aug 10 13:40:25 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3, 
> ISR=0x3, t=5015.
> Aug 10 13:40:40 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3, 
> ISR=0x3, t=1014.
> [...]
> 
> The system has three network cards.
> eth0: SIS900 (sis900.c)
> eth1: RTL-8029 (ne2k-pci.c)
> eth2: onboard VIA VT6102 Rhine-II (via-rhine.c)
> 
> eth0 and eth1 share the same interrupt
> 
>            CPU0       
>   0:      91986          XT-PIC  timer
>   1:        935          XT-PIC  keyboard
>   2:          0          XT-PIC  cascade
>   8:          1          XT-PIC  rtc
>   9:          0          XT-PIC  acpi
>  10:      25109          XT-PIC  via82cxxx, usb-uhci, usb-uhci, eth0, eth1
>  11:         24          XT-PIC  usb-uhci, eth2
>  14:       7523          XT-PIC  ide0
>  15:       7021          XT-PIC  ide1
> NMI:          0 
> ERR:          0

Wow, you have four devices on the same interrupt line. /proc/interrupts
from 2.4.26/27 looks the same?

> That was not a problem in .26 however. Though it seems to be the cause of the 
> problem (lost interrupt)? The hardware this is all running on is an Asrock 
> K7VM4 mainboard. The system is booted with "pci=noacpi" (ACPI, no APM). 
> Otherwise IRQ255 is assigned to IDE and someone told me the noacpi parameter 
> would fix the board's braindead BIOS.
> 
> Either way .27 doesn't want to boot. I've attached dmesg from a running 2.4.26 
> kernel and the config used for 2.4.27.

You mean it boots but you get the Tx timeouts?

> Other postings I've found say that the transmit timeouts mean that the 
> lowlevel ethernet connection between the NICs broke. But this works fine in 
> earlier kernels and only eth0 and eth1 which share an interrupt are affected. 
> I'd be glad for any more suggestions on what might be causing this. :)

Well there are some changes to sis900 between .26 and .27 but I doubt
they could be causing it.

Can you try to boot with ACPI disabled? I think the problem might be 
related to ACPI configuration. 

Also, can you post the boot messages from 2.4.27?

Re: eth*: transmit timed out since .27 (was: linux-2.4.27 released)

[Oliver Feiler]

[-- Attachment #1.1: body text --]
[-- Type: text/plain, Size: 4036 bytes --]

Hi Marcelo,

On Friday 13 August 2004 12:15, Marcelo Tosatti wrote:

> On Tue, Aug 10, 2004 at 02:23:34PM +0200, Oliver Feiler wrote:
> > Hi,
> >
> > I've upgraded a server from .26 to .27, but ran into problems with the
> > network cards.
> >
> > The kernel throws a lot of errors into the syslog and the net devices
> > don't work:
> > Aug 10 13:39:25 spot kernel: NETDEV WATCHDOG: eth0: transmit timed out
> > Aug 10 13:39:26 spot kernel: NETDEV WATCHDOG: eth1: transmit timed out
> > Aug 10 13:39:26 spot kernel: eth1: Transmit timeout, status 00000004
> > 00000249 Aug 10 13:39:34 spot kernel: NETDEV WATCHDOG: eth1: transmit
> > timed out Aug 10 13:39:34 spot kernel: eth1: Transmit timeout, status
> > 00000004 00000241 Aug 10 13:39:42 spot kernel: NETDEV WATCHDOG: eth1:
> > transmit timed out Aug 10 13:39:42 spot kernel: eth1: Transmit timeout,
> > status 00000004 00000240 [...]
> >
> > and:
> > Aug 10 13:39:25 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3,
> > ISR=0x3, t=515.
> > Aug 10 13:40:25 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3,
> > ISR=0x3, t=5015.
> > Aug 10 13:40:40 spot kernel: eth0: Tx timed out, lost interrupt? TSR=0x3,
> > ISR=0x3, t=1014.
> > [...]
> >
> > The system has three network cards.
> > eth0: SIS900 (sis900.c)
> > eth1: RTL-8029 (ne2k-pci.c)
> > eth2: onboard VIA VT6102 Rhine-II (via-rhine.c)
> >
> > eth0 and eth1 share the same interrupt
> >
> >            CPU0
> >   0:      91986          XT-PIC  timer
> >   1:        935          XT-PIC  keyboard
> >   2:          0          XT-PIC  cascade
> >   8:          1          XT-PIC  rtc
> >   9:          0          XT-PIC  acpi
> >  10:      25109          XT-PIC  via82cxxx, usb-uhci, usb-uhci, eth0,
> > eth1 11:         24          XT-PIC  usb-uhci, eth2
> >  14:       7523          XT-PIC  ide0
> >  15:       7021          XT-PIC  ide1
> > NMI:          0
> > ERR:          0
>
> Wow, you have four devices on the same interrupt line. /proc/interrupts
> from 2.4.26/27 looks the same?

There are five on int10. ;) It's worse on my desktop box with six devices on 
int11. But hey, Linux works just fine so I never cared.

Yes, /proc/interrupts from .26 and .27 is the same.

> > Either way .27 doesn't want to boot. I've attached dmesg from a running
> > 2.4.26 kernel and the config used for 2.4.27.
>
> You mean it boots but you get the Tx timeouts?

Yes.


> Well there are some changes to sis900 between .26 and .27 but I doubt
> they could be causing it.
>
> Can you try to boot with ACPI disabled? I think the problem might be
> related to ACPI configuration.
>
> Also, can you post the boot messages from 2.4.27?

I've attached three boots with .27. One without any parameters, one with 
acpi=off and pci=noacpi (the way I booted previous kernels).

It seems I've found the problem. The network errors were caused by the 
pci=noacpi boot parameter. Once I boot without any parameter or acpi=off it 
works just fine.

Btw, how can I boot with ACPI disabled? I thought it was acpi=off, but it 
doesn't seem to make any difference, the kernel still uses ACPI (see 
dmesg-2.4.27-acpi=off.gz attachement).

Also there must have been some (positive) changes to ACPI in 2.4.27? With 
earlier kernels I had this problem:

Feb 6 18:31:27 spot kernel: PCI: Using ACPI for IRQ routing
Feb 6 18:31:27 spot kernel: PCI: if you experience problems, try using option 
'pci=noacpi' or even 'acpi=off'
[...]
Feb 6 18:31:27 spot kernel: PCI: No IRQ known for interrupt pin A of device 
00:11.1 - using IRQ 255

This seems to have been corrected as of 2.4.27. I still get the
PCI: No IRQ known for interrupt pin A of device 00:11.1
warning, but it doesn't assign IRQ 255 anymore which I take as a good sign. :)

So, it seems to work fine now. If you still want me to test something 
regarding ACPI on this mainboard feel free to ask.

Thanks for your help,

Oliver

-- 
Oliver Feiler  -  http://kiza.kcore.de/

[-- Attachment #1.2: dmesg-2.4.27-acpi.gz --]
[-- Type: application/x-gzip, Size: 3854 bytes --]

[-- Attachment #1.3: dmesg-2.4.27-acpi=off.gz --]
[-- Type: application/x-gzip, Size: 3867 bytes --]

[-- Attachment #1.4: dmesg-2.4.27-pci=noacpi.gz --]
[-- Type: application/x-gzip, Size: 3945 bytes --]

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]