[lkml]   [2003]   [Dec]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectLinux 2.4.24-pre1


Here goes 2.4.24-pre1...

The XFS filesystem has been merged.

This release contains mostly architecture specific updates.

Summary of changes from v2.4.23 to v2.4.24-pre1

o ia64: make hpzx1 fake pci device safer

o ac97_plugin_ad1980 fixes

o ia64: Remove platform_pci_enable_device() machine vector and synchronize sba_iommu.c with 2.5.
o ia64: Bail out of SBA init function if no IOC found. Avoids spurious (but harmless) "No IOC for PCI Bus 0000:00 in ACPI" messages when booting generic kernel on non-ZX1 hardware.
o ia64: Clear corrected errors (CMCs and CPEs) in the kernel
o ia64: The "HP_ZX1" kernel works on sx1000-based machines as well as zx1-based ones, so make the descriptions a little more generic.
o ia64: add kmap_types.h to make crypto, etc compile. (This is just a dummy file from 2.6 and shouldn't ever be used.)
o ia64: fix EFI memory map trimming
o Cset exclude:[helgaas]|ChangeSet|20031030215302|13517
o ia64: update default configs

o Gone dmapi

o [XFS] switch xfs to use linux imode flags internally

o Import changeset

o [SERIAL] Make the Startech & 16552D UART detection 'more correct'

o ia64: put xor functions in .S file (backported from 2.6)

o ia64: fix show_mem() panic

o [NET]: Allow SOMAXCONN to be adjusted via sysctl

o ia64: Don't print anything for unimplemented syscalls

o Ernie Petrides: Readd exec_mmap() fastpath with correct locking
o Readd the OOM killer as configurable option, defaulted to off

o [XFS] Remove some unused pagebuf source and header files
o Fix utimes(2) and immutable/append-only files. Originally by Ethan Benson
o Remove some unused macros and related comment from the XFS quota header
o Add a process flag to identify a process performing a transaction
o Support for delayed allocation. Used by XFS and backported from 2.6
o Provide a simple try-lock based dirty page flushing routine
o Provide an iget variant without unlocking the inode and without the read_inode call (iget_locked). Used by XFS and backported from 2.6.
o Export several kernel symbols used by the XFS filesystem
o Add XFS documentation and incorporate XFS into the kernel build


o 2.4 lacks dummy SET_NETDEV_DEV

o [XFS] pv=892598;; Change xlog_verify_iclog() to use idx as zero based instead

o [XFS] Work around gcc 2.96 bug in _lsn_cmp

o [TG3]: Add new device IDs

Adam Radford:
o 3ware driver update for 2.4.23-bk2

Arun Sharma:
o ia64: MINSIGSTKSZ on ia32
o ia64: CONFIG_IA32_SUPPORT can only be static, not a module
o ia64: make strace of ia32 processes work again
o ia64: Don't mix user/kernel pointers in 32-bit stat/statfs emulation

Ben Greear:

Chris Mason:
o From -aa tree: Fix end_buffer_io_kiobuf() locking

Chuck Lever:
o Make readahead last page of file

David Hinds:
o update/bugfix for pcnet_cs driver

David Mosberger:
o ia64: In <asm-ia64/param.h>, do not include <linux/config.h> outside the #ifdef __KERNEL__ bracket. Doing so pollutes the user- level namespace. Bug report & proposed fix by GOTO Masanori.
o ia64: Control /proc/bus/mckinley/zx1 via separate SBA_PROC_FS macro and turn SBA_PROC_FS off by default (it's too much of a scalability bottleneck).
o ia64: Mark access_ok() as likely to succeed (as is done in x86 tree)
o ia64: Fix efi_mem_type() and efi_mem_attributes() to avoid potential underflows. In my case, the underflows occurred with the first memory descriptor which got trimmed down to a size of 0.
o ia64: Fix a alternate-signal-stack bug which could corrupt RNaT bits when bspstore happened to point to an RNaT-slot.
o ia64: Fix a bug in sigtramp() which corrupted ar.rnat when unwinding across a signal trampoline (in user space). Reported by Laurent Morichetti.

David S. Miller:
o [TCP]: Put Alexey's -EAGAIN change back in with Linus's fix on top
o [NETLINK]: Initialize nl_pad in getname and recvmsg, noticed by Uli Drepper
o [PACKET]: In packet_recvmsg(), test correct flags for MSG_TRUNC handling
o [PACKET]: Revert MSG_TRUNC change, the original behavior was intentional
o [NET]: In sock_queue_rcv_skb(), do not deref skb->len after it is queued to the socket
o [PPPOE]: Do not leak SKB if sock_queue_rcv_skb() fails
o [ECONET]: Do not leak SKBs if ec_queue_packet() fails
o [SPARC64]: Fix non-modular build of FFB drm driver
o [VLAN]: Kill build warning due to missing declarations
o [TG3]: Do not drop existing GRC_MODE_HOST_STACKUP when writing to GRC_MODE
o [TG3]: Do not set RX_MODE_KEEP_VLAN_TAG when ASF is enabled
o [TG3]: Clear on-chip stats/status block after resetting flow-through queues
o [TG3]: Update version and release date
o [TG3]: Update to latest non-5705 TSO firmware

David Stevens:
o [IPV6]: Fix UDP socket selection for multicast
o [IPV6]: Fix milliseconds to jiffies conversion in multicast code
o [IPV6]: In multicast code, set MAF_TIMER_RUNNING when timer is set
o [IPV6]: In igmp6_group_queried, fix address check to comply with RFC2710

Dean Roehrich:
o [XFS] fix some ia64 warnings in dmapi_xfs.c
o [XFS] Change dm_send_namesp_event to take vnode ptrs rather than bhv ptrs
o [XFS] Change dm_send_mount_event to use vnode ptrs rather than bhv ptrs
o [XFS] Change dm_send_destroy_event to use vnode ptrs rather than bhv ptrs
o [XFS] Make dm_send_data_event use vp rather than bhv
o [XFS] Implement dm_get_bulkall
o [XFS] Remove duplicate FILP_DELAY_FLAG macro
o [XFS] dm_path_to_handle returns errnos with sign flipped

Eric Sandeen:
o [XFS] Re-work xfs stats macros to support per-cpu data
o [XFS] remove doubly-included header files
o [XFS] Re-work pagebuf stats macros to help support per-cpu data
o [XFS] Update sysctls - use ints, not ulongs, and show pagebuf values in jiffies like everybody else
o [XFS] Allow full 32 bits in sector number when XFS_BIG_BLKNOS not set
o [XFS] Add a stack trace to _xfs_force_shutdown
o [XFS] Fix test for large sector_t when finding max file offset
o [XFS] Use i_size_read/i_size_write semantics from 2.6 kernel to reduce 2.4/2.6 differences in xfs
o [XFS] Use buffer head flag set/clear routines as in 2.6 kernel to reduce 2.4/2.6 differences in xfs
o [XFS] Remove a nested transaction in xfs_dm_punch_hole
o [XFS] BH_Sync added in 2.4.22, put an #ifdef in for now so this still works on older kernels.
o [XFS] Fix a few sysctls - values are all ints, but sysctl table was setting up for longs.
o [XFS] Fix the pb stats clear handler, value is int but handler was using ulong

Geert Uytterhoeven:
o [NET]: Fix atm/br2684 build with procfs disabled
o Atyfb on Mach64 GX or Atari
o 2.4.23 ext3 warning
o M68k RMW accesses
o Zorro include guard
o M68k cache mode
o Mac ADB IOP fix
o M68k symbol exports
o Mac89x0 Ethernet
o Mac ESP SCSI setup
o Macfb setup
o Mac ADB
o ncr53c7xx SCSI
o Amiga debug fix
o Amiga Gayle E-Matrix 530 IDE
o Amiga NCR53c710 SCSI
o Mac II VIA
o M68k asm/system.h
o M68k extern inline
o M68k Documentation
o Amiga Buddha/CatWeasel IDE
o NCR53C9x SCSI inline
o Genrtc warning
o Mac SCSI
o Mac extern
o Amiga Gayle IDE cleanup

Glen Overby:
o [XFS] A problem was found with the debug code in xlog_state_do_callback. At the end of processing all log buffers that can be processed, there is a (debug only) double-check to make sure that log buffers with completed I/O don't get marooned when the function completes. The check only needs to go to the first buffer that will cause an I/O completion, that has not completed. The loop doesn't stop a WANT_SYNC state buffer is found, but it should.
o [XFS] xfs_dir2_node_addname_int had reminants of an old block placement algorithm in it. The old algorithm appeared to look for the first place to put a new data block, and thus a new freespace block (this is where the 'foundindex' variable came from). However, new space in a directory is always added at the lowest file offset as determined by the extent list. So this stuff is never used.

Harald Welte:
o [NETFILTER]: Fix ip_queue_maxlen sysctl
o [NETFILTER]: Kill extraneous memset()s in nat/conntrack TFTP code
o [NETFILTER]: Fix ipchains MASQUERADE oops
o [IPV6]: Do not bypass netfilter for MLD/IGMP messages

Herbert Xu:
o [SCTP]: Fix sm.h/sctp.h header include loop

Hideaki Yoshifuji:
o [CRYPTO]: crypto_alg_lookup() should fail when passed a NULL name

Jesse Barnes:
o ia64: protect PAL mapping printk with EFI_DEBUG

John Stultz:
o Always make TSC available via get_cycles() when TSC is present

Jozsef Kadlecsik:
o [NETFILTER]: Use list macros instead of reimplementation
o [NETFILTER]: Avoid order n^2 lookup of whole conntrack hash in ip_ct_selective_cleanup()

Keith Owens:
o ia64: fix offsets.h generation bootstrap problem
o ia64: mca_asm.h documentation fixes
o ia64: Trivial fixes for correct field type in formats. prfunc_t does not include attribute format so gcc does not pick these up automatically.
o ia64: salinfo.c cleanup and race removal
o ia64: print header from INIT records
o ia64: fix comment typo (sal.h)
o ia64: Clean up kernel salinfo state checking
o ia64: Add the ability for user space salinfo to ask kernel salinfo and/or the prom to decode the oem data sections of SAL records.
o ia64: Update PAL_MC_ERROR_INFO structures for SDM 2.1
o ia64: sync salinfo.c with 2.6 (suser -> capable, use standard macros)

Khalid Aziz:
o ia64: do_settimeofday: fix compensation for lost ticks

Kochi Takayoshi:
o ia64: initialize bootmem later, since acpi_table_init() doesn't need it
o ia64: don't access per-CPU data of off-line CPUs

Krishna Kumar:
o [IPV6]: Fix ref count bug in MLDv2, test idec->dead instead of IFF_UP

Linus Torvalds:
o Fix x86 kernel page fault error codes

Matt Domsch:
o EDD: read disk80 MBR signature, export through edd module

Matthew Wilcox:
o Remove broken file lock accounting

Mikael Pettersson:
o fix some DRM43 warnings
o fix reboot/no_idt bug

Mikael Starvik:
o CRIS architecture update

Mikulas Patocka:
o from -aa tree: Fix potential fsync() race condition

Nathan Scott:
o [XFS] Fix up the default ACL inherit case, in the presence of failure during applying the default ACL (eg. from ENOSPC)
o [XFS] Fix a compiler warning, sync_fs returns a value
o [XFS] Fix a race condition in async pagebuf IO completion, by moving blk queue manipulation down into pagebuf. Fix some busted comments in page_buf.h, use a more descriptive name for __pagebuf_iorequest
o [XFS] Use the rounded down size value for all growfs calculations, else the last AG can be updated incorrectly
o [XFS] Fix a harmless typo - we were using a pagebuf flag not a bmap flag here; fortunately they have the same value (2)
o [XFS] Tweak last dabuf fix, suggested by Steve, no longer uses bitfields but uchars instead
o [XFS] Fix a case where we could issue an unwritten extent buffer for IO without it being locked - an instant BUG trigger in the block layer
o [XFS] Alternate, cleaner fix for the ENOSPC/ACL lookup problem
o [XFS] Automatically set logbsize for larger stripe units
o [XFS] Add inode64 mount option; fix case where growfs can push 32 bit inodes into 64 bit space accidentally - both changes originally from IRIX
o [XFS] Fix races between O_DIRECT and fcntl with F_SETFL flag on the XFS IO path
o [XFS] DMAPI changes required by direct IO/fcntl setfl interaction races
o [XFS] Separate the big filesystems macro out into separate big inums and blknos macros. Also fix the check for too-large filesystems in the process
o [XFS] Undo last mod, checked in against wrong bug number with wrong change message
o [XFS] Separate the big filesystems macro out into separate big inums and blknos macros; fix the check for too-large filesystems in the process
o [XFS] Implement several additional inode flags - immutable, append-only, etc; contributed by Ethan Benson
o [XFS] Some tweaks to the additional inode flags, suggested by Ethan
o [XFS] Accidentally switched some debug code off, reenable it
o [XFS] Allow syncing the types header up more easily with userspace
o [XFS] Make debug code _exactly_ how it used to be to save on tree merging
o [XFS] Change writepage code so that we mark a page uptodate if all of its buffers are uptodate, and we are not doing a partial page write
o [XFS] Remove xfs_attr_fetch.c - the one routine was a copy of another, so instead of fixing a bug in two places I merged the two routines
o [XFS] Use the same name for a function here as in the 2.5/2.6 tree
o [XFS] Use xfs_dev_t size rather than dev_t size in xfs_attr_fork initialization
o [XFS] Rename _inode_init_once to __inode_init_once to follow the kernel naming convention a bit more closely
o [XFS] Fix a broken interaction between a buffered read into an unwritten extent and a direct write
o [XFS] Clean up inode revalidation code slightly
o [XFS] Fix up pointers in diagnostics, print using %p not %x for 64 bit platforms
o [XFS] Rename pagebuf debug option (ie. pagebuf tracing) into a generic XFS tracing option for the other XFS trace code to use too (once fixed)
o [XFS] Fix compiler warning after change to xfs_ioctl interface
o [XFS] Use an xfs_ino_t to hold the result of inode extraction from a handle, not a possibly 32-bit number
o [XFS] final round of code cleanup, now using 3-clause-bsd in these headers
o [XFS] Fix inode btree lookup code precision problem with large allocation groups
o [XFS] Add some IO path tracing, move inval_cached_pages to a better home to help
o [XFS] Fix ktrace code - dont build unilaterally, and do earlier init for pagebuf use
o [XFS] Fix log tracing code so it is independent of DEBUG like other traces
o [XFS] Add back xfsidbg tracing code, remove ktrace<->debug dependency
o [XFS] Fix build fallout from reordering xfsidbg headers for tracing fixes
o [XFS] Rename the vnode tracing macro to be consistent with the other trace code
o [XFS] Enable tracing in the quota code if requested
o [XFS] Fix exports for tracing symbol access in idbg code
o [XFS] When tracing extended attribute calls, only access the buffer when it exists
o [XFS] Fix build with tracing enabled, couple of portability macros, move externs into headers
o [XFS] Enable the tracing options in XFS Makefiles
o [XFS] Fix compile warning on 64 bit platforms
o [XFS] Fix compiler warning when building on 2.4.21 kernels
o [XFS] Dont build objects which are not linked into the kernel ever
o [XFS] Dump the pagebuf locked field for debugging purposes
o [XFS] Fix warnings when tracing enabled on 64 bit platforms
o [XFS] Fix pagebuf page locking problems for blocksizes smaller than the pagesize
o [XFS] Fix a supplemental issue introduced by the last small blocksize locking fix; this would manifest itself as a second unlock_page call on an already unlocked page
o [XFS] Fix a deadlock while writing when low on free space
o [XFS] Remove some spurious 2.4/2.6 differences in support code
o [XFS] Fix sign on a pagebuf error variable, backport from 2.6 tree
o [XFS] Fix an infinite writepage loop under a combination of low free space, and racing write/unlink calls to the same file
o [XFS] Enable pagebuf lock tracking via debug
o [XFS] Backport a couple of debugging changes from the 2.6 code base
o [XFS] Backport minor 2.6 changes to the iomap interface to keep code more in sync
o [XFS] Backport an unmerged bug fix from the 2.6 code base - if probe_unmapped_page fails while walking down the unmapped page list, do not attempt to probe the last page as well just return
o [XFS] Backport an unmerged bug fix from the 2.6 code base - only submit a convert_page page for IO if startio is set
o [XFS] Backport some trivial changes from the 2.6 code base - page uptodate flag macro name changes
o [XFS] Move Linux-version specific code out of xfs_iomap.c so that it can become part of the XFS core code
o [XFS] Seperate the NFS reference cache code out from xfs_rw.c to simplify management of different kernel versions
o [XFS] Remove assertion that we do not hold a lock - no lock ownership state available
o [XFS] Merge page_buf_locking routines in with the rest of page_buf
o [XFS] Change pagebuf to use the same ktrace implementation as XFS, instead of reinventing that wheel
o [XFS] Trivial/whitespace changes to sync up different trees a bit
o [XFS] Switch to using the BSD qsort implementation
o [XFS] Fix a build error in some debug code
o [XFS] Fix build fallout from refcache reorganisation
o [XFS] Move the stack trace wrapper into a kernel-version-specific location
o [XFS] Switch from using dev_t to xfs_buftarg_t for representing the devices underneath XFS
o [XFS] Merge find_next_zero_bit casting fixes back from 2.6 code
o [XFS] Use iomap abstraction consistently
o [XFS] Abstract sendfile operation out, supporting multiple kernels more easily
o [XFS] Use xfs_statfs type to statfs operation, to support multiple kernels more easily
o [XFS] Switch debug quota code to use xfs_buftarg interface instead of dev_t
o [XFS] Abstract out the current_time interface use from quota to support multiple kernel versions
o [XFS] Fix some incorrect debug code after buftarg changes
o [XFS] Use a kmem shaking interface for 2.4 which is much more like the 2.6 one
o [XFS] Convert to revised kmem shake interface
o [XFS] Update the way we hook into the generic direct IO code so we share more code. This means we no longer need to dup remove_suid within xfs_write_clear_setuid
o [XFS] Add the noikeep mount option, make ikeep the default for now
o [XFS] Use vnode timespec modifiers for atime/mtime/ctime, keeps last code hunk in sync
o [XFS] Prevent log ktrace code from sleeping in an invalid context
o [XFS] Fix comment in xfs_rename.c

Neil Brown:
o Make root a special case for per-user process limits
o Honour SUN NFSv2 hack for "set times to server time
o Drop module count if lockd reclaimer thread failed to start

Patrick McHardy:
o [NET SCHED]: Adjust qlen when grafting in multiple qdiscs
o [NET SCHED]: Reset q.qlen in tbf_reset instead of purging an unused queue
o [NET SCHED]: Fix queue limits in multiple qdiscs
o [NETFILTER]: Fix various issues with the amanda conntrack+NAT helpers
o [NETFILTER]: Fix expectation evict order
o [NETFILTER]: Export conntrack bucket count via read-only sysctl

Ralf Bächle:
o MIPS updates
o Update MIPS MAINTAINERS entries
o Update generic MIPS code
o mips64 updates
o AMD Alchemy updates
o Remove support for compressed EV-64120 kernels
o Add support for Momentum Ocelot C and Jaguar ATX
o Add VINO and Indycam drivers
o Update SGI IP22 support
o Update RM200 C support
o Update support for Lasat boards
o Update for NEC VR41xx systems
o Update TX4927 code
o MIPS char driver update
o Update Origin code
o MIPS video driver updates
o Update Atlas, Malta and SEAD boards
o Update sgiwd93 driver
o Sibyte updates
o Update JMR3927 support
o SGI I2C driver
o NEC DDB updates
o ITE updates
o Galileo boards
o DEC updates
o Update Jazz
o Update HP Laserjet
o Update Ocelot defconfig
o Update Cobalt defconfig file
o Update Philips Nino defconfig
o ARC library updates
o Update NEC Osprey defconfig
o Add GT-064011/GT-64111 PCI ID
o Fix VRc5477 remove method declaration
o Add two new NEC PCI IDs
o Probe SGI partitions earlier
o Add Pete Popov to CREDITS
o Add IP22 parallel port driver
o New PCMCIA drivers
o Include <asm/system.h> into spinlock.h
o Add more MIPS bits in <linux/elf.h>
o MIPS ioaddr_t is 32-bit
o Remove remaining drivers/sgi bits
o Cleanup lk201-map.c
o No stone-age compat stuff for MIPS
o Supply default values to rtc.c
o Misc MIPS video bits
o MIPS network driver updates

Russell Cattelan:
o [XFS] Fix some inconsistent types
o [XFS] Rework pagebuf_delwri_flush to be list safe
o [XFS] Fix one more fsid_t type
o [XFS] Clean up fsid_t abuses in dmapi
o [XFS] Since we now have embeding trees and XFS has to support LBS which typically 1 version back from the XFS TOT tree add support for 2.4.22 with and #if KERNEL_VERSION
o [XFS] Fix from Christoph
o [XFS] IRIX sets KM_SLEEP to 0 but the support routines sets KM_SLEEP to 1
o [XFS] Fix remount,ro path
o [XFS] move the iomap data structures out of pagebuf
o [XFS] Add new file ... missed in orginal checkin

Rusty Russell:
o [NETFILTER]: Do not flush MASQ if IP did not change

Stephen Hemminger:
o [NETFILTER]: Trivial -- Get rid of warnings in netfilter if /proc is not configured on

Stephen Lord:
o [XFS] do not put 0x in front of a decimal number, its confusing
o [XFS] fix up xfs_lowbit's use of ffs
o [XFS] fix build for gcc 3.2
o [XFS] Make xfs_ichgtime call mark_inode_dirty_sync instead of mark_inode_dirty makes the just the inode look dirty, and not the inode and the data.
o [XFS] remove an impossible code path from mkdir and link paths, spotted by Al Viro.
o [XFS] Switch pagebuf hashing to be based on the block_device address rather than the dev_t. Should give better distribution.
o [XFS] remove dead function xfs_trans_iput
o [XFS] Close some holes in the metadata flush logic used during unmount, make sure we have no pending I/O completion calls for metadata, and that we only keep hold of metadata buffers for I/O completion if we want to. Still not perfect, but better than it was.
o [XFS] When calculating the number of pages to probe for an unwritten extent, use the size of the extent, not the page count of the pagebuf which is initialized to zero.
o [XFS] Rework how xfs and the linux generic I/O code interoperate again to deal with deadlock issues between the i_sem and i_alloc_sem and the xfs IO lock.
o [XFS] move unwritten extent conversion for O_DIRECT into the write thread and out of the I/O completion threads. This scales better.
o [XFS] Code cleanup
o [XFS] small cleanup
o [XFS] fix the previous change which compiled by fluke, the conditional use of the i_alloc_sem was wrong. No actual change in the generated code for 2.4.22, there will be for older kernels though.
o [XFS] fix up error unlock paths in xfs_write
o [XFS] Implement deletion of inode clusters in XFS
o [XFS] cleanup uio use some more
o [XFS] remove FINVIS from xfs, instead use a seperate file ops vector for files which are opened for invisible I/O.

Steven Cole:
o 2.4.23 update Documentation/Changes for quota-tools

Stéphane Eranian:
o ia64: perfmon-1 inheritance bugfix

Tom Rini:
o PPC32: Convert all bootwrappers that use OpenFirmware to use the same code
o PPC32: Allow for the commandline to be pulled from OF on PReP
o PPC32: Backport some warning fixes to arch/ppc/boot/prep/vreset.c
o PPC32: Finish support for pinning TLB entries on MPC8xx
o PPC32: Workaround some errata on the MPC74xx line
o PPC32: Updates for the IBM 750FX processor
o PPC32: Don't fudge the MAC address on EP8260's
o PPC32: Add a cputable entry for the Motorola MPC8280
o PPC32: gcc-3.4 build fixes from Olaf Hering <>
o PPC32: Print the correct ammount of memory not covered by BATs
o PPC32: Update an errata on the MPC745x line
o Fix PPC compilation

Tony Luck:
o ia64: infinite loop in ia64_mca_wakeup_ipi_wait
o ia64: fix register numbers in MCA save/restore
o ia64: Another MCA fix

Ville Nuorvala:
o [IPV6]: Verify nlmsg_len in rt6_dump_route()

William Lee Irwin III:
o Fixup smb_boot_cpus(): Fix HT detection bug
o out_of_memory() locking
o fix 2.4 BLK_BOUNCE_ANY

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:59    [W:0.080 / U:0.708 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site