]> git.neil.brown.name Git - history.git/log
history.git
23 years ago[PATCH] The epoll saga continues ...
Davide Libenzi [Fri, 8 Nov 2002 03:54:55 +0000 (19:54 -0800)]
[PATCH] The epoll saga continues ...

Proper wakeup code in ep_insert and ep_modify

23 years ago[PATCH] Fix ALSA emu10k1 bass control
Rob Weryk [Fri, 8 Nov 2002 03:25:49 +0000 (19:25 -0800)]
[PATCH] Fix ALSA emu10k1 bass control

This trivial patch fixes a mixer problem with the emu10k1 driver in
ALSA.  In sound/pci/emu10k1/emufx.c, the line

static const u32 bass_table[41][5] = {

only has 40 lines defined, instead of 41.  This results in no sound
output when the bass control is set at 100% (but works fine at 98%) I
added the missing line, which is present in the OSS emu10k1 driver.

23 years ago[PATCH] let binfmt_misc optionally preserve argv[1]
David Mosberger [Fri, 8 Nov 2002 03:23:22 +0000 (19:23 -0800)]
[PATCH] let binfmt_misc optionally preserve argv[1]

This makes it possible for binfmt_misc to optionally preserve the
contents of argv[1].  This is needed for building accurate simulators
which are invoked via binfmt_misc.  I had brought up this patch a while
ago (see URL below) and there was no negative feedback (OK, there was no
feedback at all...  ;-).

The patch is trivial and the new behavior is triggered only if the
letter "P" (for "preserve") is appended to the binfmt_misc registration
string, so it shold be completely safe.

23 years ago[PATCH] C99 designated initializers for drivers/char
Art Haas [Fri, 8 Nov 2002 03:22:02 +0000 (19:22 -0800)]
[PATCH] C99 designated initializers for drivers/char

23 years ago[PATCH] C99 designated initializer for include/linux/cpufreq.h
Art Haas [Fri, 8 Nov 2002 03:19:17 +0000 (19:19 -0800)]
[PATCH] C99 designated initializer for include/linux/cpufreq.h

23 years ago[PATCH] C99 designated initializers for fs/fat
Art Haas [Fri, 8 Nov 2002 03:19:07 +0000 (19:19 -0800)]
[PATCH] C99 designated initializers for fs/fat

23 years ago[PATCH] C99 designated initializers for fs/umsdos
Art Haas [Fri, 8 Nov 2002 03:18:57 +0000 (19:18 -0800)]
[PATCH] C99 designated initializers for fs/umsdos

23 years ago[PATCH] C99 designated initializers for fs/affs
Art Haas [Fri, 8 Nov 2002 03:18:46 +0000 (19:18 -0800)]
[PATCH] C99 designated initializers for fs/affs

23 years ago[PATCH] C99 designated initializer for kernel/cpufreq.c
Art Haas [Fri, 8 Nov 2002 03:18:36 +0000 (19:18 -0800)]
[PATCH] C99 designated initializer for kernel/cpufreq.c

23 years agoMerge master.kernel.org:/home/davem/BK/net-2.5
Linus Torvalds [Fri, 8 Nov 2002 01:16:18 +0000 (17:16 -0800)]
Merge master.kernel.org:/home/davem/BK/net-2.5
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoFrom Rick Lindsley <ricklind@us.ibm.com>: missing return value
Linus Torvalds [Fri, 8 Nov 2002 01:14:40 +0000 (17:14 -0800)]
From Rick Lindsley <ricklind@us.ibm.com>: missing return value
in sysfs partition code.

23 years ago[IPSEC]: Fix lockup in xfrm4_dst_check.
Alexey Kuznetsov [Thu, 7 Nov 2002 23:09:35 +0000 (15:09 -0800)]
[IPSEC]: Fix lockup in xfrm4_dst_check.

23 years ago[NETFILTER]: Remove duplicate include.
Brad Hards [Thu, 7 Nov 2002 17:12:47 +0000 (09:12 -0800)]
[NETFILTER]: Remove duplicate include.

23 years ago[SCTP]: Remove duplicate include.
Brad Hards [Thu, 7 Nov 2002 17:11:52 +0000 (09:11 -0800)]
[SCTP]: Remove duplicate include.

23 years agoMerge nuts.ninka.net:/home/davem/src/BK/network-2.5
David S. Miller [Thu, 7 Nov 2002 16:40:11 +0000 (08:40 -0800)]
Merge nuts.ninka.net:/home/davem/src/BK/network-2.5
into nuts.ninka.net:/home/davem/src/BK/net-2.5

23 years ago[IPSEC] More work.
Alexey Kuznetsov [Thu, 7 Nov 2002 16:36:15 +0000 (08:36 -0800)]
[IPSEC] More work.
1. Expiration of SAs. Some missing updates of counters.
   Question: very strange, rfc defines use_time as time of the first use
   of SA. But kame setkey refers to this as lastuse.
2. Bug fixes for tunnel mode and forwarding.
3. Fix bugs in per-socket policy: policy entries do not leak but are destroyed,
   when socket is closed, and are cloned on children of listening sockets.
4. Implemented use policy: i.e. use ipsec if a SA is available,
   ignore if it is not.
5. Added sysctl to disable in/out policy on some devices.
   It is set on loopback by default.
6. Remove resolved reference from template. It is not used,
   but pollutes code.
7. Added all the SASTATEs, now they make sense.

23 years ago[PATCH] read(v)/write(v) fixes
Christoph Hellwig [Thu, 7 Nov 2002 15:57:06 +0000 (07:57 -0800)]
[PATCH] read(v)/write(v) fixes

Clean up vfs_readv/writev() interface and avoid code duplication.

Make kNFSd use the cleaned-up interfaces, and disable the code that
accesses the low-level readpage() function of the exported filesystem
(not allowed - many filesystems need extra setup, which is why we have
a separate ->sendpage() routine for that).

23 years ago[PATCH] CONFIG_STACK_GROWSUP
Matthew Wilcox [Thu, 7 Nov 2002 08:56:27 +0000 (00:56 -0800)]
[PATCH] CONFIG_STACK_GROWSUP

Change ARCH_STACK_GROWSUP to CONFIG_STACK_GROWSUP as requested.

23 years ago[PATCH] Broken Hugetlbpage support in 2.5.46
Seth Rohit [Thu, 7 Nov 2002 08:46:19 +0000 (00:46 -0800)]
[PATCH] Broken Hugetlbpage support in 2.5.46

The hugetlb page support in 2.5.46 is broken (Don't know if this is the
first version of kernel or any prior revs also have that).  Basically the
free side of hugepages was really not freeing the physical resources (for
the cases when the pages allocated using system call interface).  Attached
is the patch that should resolve it. (doesn't break the hugetlbfs support
either).

23 years ago[PATCH] C99 designated initializers for fs/hugetlbfs/inode.c
Art Haas [Thu, 7 Nov 2002 08:46:13 +0000 (00:46 -0800)]
[PATCH] C99 designated initializers for fs/hugetlbfs/inode.c

23 years ago[PATCH] bug in NFSv2 end-of-file read handling
Chuck Lever [Thu, 7 Nov 2002 08:43:38 +0000 (00:43 -0800)]
[PATCH] bug in NFSv2 end-of-file read handling

NFSv2 doesn't pass connectathon 2002, at least on some Linux kernels.
Trond deemed the following modification necessary in all kernels to
address the problem.

23 years ago[PATCH] too many setattr calls from VFS layer
Chuck Lever [Thu, 7 Nov 2002 08:43:31 +0000 (00:43 -0800)]
[PATCH] too many setattr calls from VFS layer

New code in 2.5 VFS layer invokes notify_change to clear the suid and
sgid bits for every write request.  notify_change needs to optimize out
calls to ->setattr that don't do anything, because for many network file
systems, an on-the-wire SETATTR request is generated for every ->setattr
call, resulting in unnecessary latency for NFS writes.

23 years ago[PATCH] allow nfsroot to mount with TCP
Chuck Lever [Thu, 7 Nov 2002 08:43:25 +0000 (00:43 -0800)]
[PATCH] allow nfsroot to mount with TCP

nfsroot needs to pass the network protocol (UDP/TCP) into the mount
functions in order to support mounting root partitions via NFS over TCP.

23 years ago[PATCH] kconfig documentation update
Roman Zippel [Thu, 7 Nov 2002 08:41:01 +0000 (00:41 -0800)]
[PATCH] kconfig documentation update

This removes the old documentation, adds the new one and fixes all
references to it.

23 years ago[PATCH] various kconfig updates
Roman Zippel [Thu, 7 Nov 2002 08:40:45 +0000 (00:40 -0800)]
[PATCH] various kconfig updates

Various small kconfig updates to fix all the reported little problems and
the single menu mode for menuconfig by Petr Baudis <pasky@ucw.cz>.

23 years ago[PATCH] remove old config tools
Roman Zippel [Thu, 7 Nov 2002 08:39:40 +0000 (00:39 -0800)]
[PATCH] remove old config tools

This deletes the old config tools and moves Michael's maintainer entry
for them to CREDITS and I added myself for KCONFIG instead.

23 years agoMerge http://gkernel.bkbits.net/net-drivers-2.5
Linus Torvalds [Thu, 7 Nov 2002 07:11:28 +0000 (23:11 -0800)]
Merge http://gkernel.bkbits.net/net-drivers-2.5
into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

23 years agoProperly terminate b44 net driver's PCI id table
Jeff Garzik [Thu, 7 Nov 2002 09:53:07 +0000 (04:53 -0500)]
Properly terminate b44 net driver's PCI id table
(caught by Arjan @ Red Hat)

23 years agoUse dev_kfree_skb_any not dev_kfree_skb in tg3 net driver
Jeff Garzik [Thu, 7 Nov 2002 09:03:26 +0000 (04:03 -0500)]
Use dev_kfree_skb_any not dev_kfree_skb in tg3 net driver
function tg3_free_rings.

Spotted by DaveM.

23 years agoMore znet net driver updates. Driver now survives plug/unplug of cable.
Marc Zyngier [Thu, 7 Nov 2002 08:49:53 +0000 (03:49 -0500)]
More znet net driver updates.  Driver now survives plug/unplug of cable.

23 years agoConvert 3c505 net driver to use spinlocks instead of cli/sti
Paulo André [Thu, 7 Nov 2002 08:35:32 +0000 (03:35 -0500)]
Convert 3c505 net driver to use spinlocks instead of cli/sti

23 years agoThe patch below contains some minor updates to the bmac and mace
Paul Mackerras [Thu, 7 Nov 2002 08:18:22 +0000 (03:18 -0500)]
The patch below contains some minor updates to the bmac and mace
ethernet drivers used on powermacs.  The bmac.c change is just to
remove some compile warnings.  The mace.c change is to move an inline
function definition to before the point where it is used.

23 years agoMerge DaveM's cleanup of Broadcom's GPL'd 4401 net driver
Jeff Garzik [Thu, 7 Nov 2002 08:07:09 +0000 (03:07 -0500)]
Merge DaveM's cleanup of Broadcom's GPL'd 4401 net driver

23 years agoMerge redhat.com:/home/jgarzik/repo/linus-2.5
Jeff Garzik [Thu, 7 Nov 2002 07:32:15 +0000 (02:32 -0500)]
Merge redhat.com:/home/jgarzik/repo/linus-2.5
into redhat.com:/home/jgarzik/repo/net-drivers-2.5

23 years agoMerge http://gkernel.bkbits.net/irda-2.5
Linus Torvalds [Thu, 7 Nov 2002 05:53:49 +0000 (21:53 -0800)]
Merge http://gkernel.bkbits.net/irda-2.5
into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

23 years agoIrDA update 3/3:
Jean Tourrilhes [Thu, 7 Nov 2002 10:58:54 +0000 (05:58 -0500)]
IrDA update 3/3:
<Thanks to Martin Diehl>
o [CORRECT] Handle non-linear and shared skbs
o [CORRECT] Tell kernel we can handle multithreaded receive
<Of course, this has been tested extensively on SMP>

23 years agoIrDA update 2/3:
Jeff Garzik [Thu, 7 Nov 2002 10:57:21 +0000 (05:57 -0500)]
IrDA update 2/3:
(Adrian Bunk)
* C99 initializers
* fix public symbol name conflict
(me)
* further clean up namespace on donauboe driver in module_init/exit area

23 years agoIrDA update 1/3:
Jean Tourrilhes [Thu, 7 Nov 2002 10:35:25 +0000 (05:35 -0500)]
IrDA update 1/3:
        <Following patch from Martin Diehl>
o [CRITICA] Do all serial driver config change in process context
o [CORRECT] Safe registration of dongle drivers
o [FEATURE] Rework infrastructure of SIR drivers
o [CORRECT] Port irtty driver to new SIR infrastructure
o [CORRECT] Port esi/actisys/tekram driver to new SIR infrastructure
<Note : there is still some more work to do around SIR drivers,
 such as porting other drivers to the new infrastructure, but
 this is functional and tested, and old irtty is broken>

23 years agoMerge nuts.ninka.net:/home/davem/src/BK/sctp-2.5
David S. Miller [Thu, 7 Nov 2002 01:06:01 +0000 (17:06 -0800)]
Merge nuts.ninka.net:/home/davem/src/BK/sctp-2.5
into nuts.ninka.net:/home/davem/src/BK/net-2.5

23 years ago[PATCH] support non-rt signals >32
Matthew Wilcox [Thu, 7 Nov 2002 00:13:07 +0000 (16:13 -0800)]
[PATCH] support non-rt signals >32

On PA-RISC, SIGRTMIN is 36, so a 32-bit data type is not enough.  We
conditionalise it so other arches don't pay the cost.

23 years ago[PATCH] generic prefetch support in xor.h
Matthew Wilcox [Thu, 7 Nov 2002 00:13:02 +0000 (16:13 -0800)]
[PATCH] generic prefetch support in xor.h

Add prefetching support to asm-generic/xor.h.  This gives a healthy speedup
on both PA-RISC & IA64.

23 years ago[PATCH] misc updates
Matthew Wilcox [Thu, 7 Nov 2002 00:12:57 +0000 (16:12 -0800)]
[PATCH] misc updates

- CREDITS & MAINTAINERS updates
- changes for the new kstat/dkstat struct
- Kconfig updates
- L_TARGET isn't obsolete yet
- fix the sys_truncate/truncate64 issue properly this time
- add MAP_POPULATE & MAP_NONBLOCK definitions

23 years ago[PATCH] initramfs support
Matthew Wilcox [Thu, 7 Nov 2002 00:12:52 +0000 (16:12 -0800)]
[PATCH] initramfs support

Support initramfs on parisc

23 years ago[PATCH] C99 initialisers
Matthew Wilcox [Thu, 7 Nov 2002 00:12:48 +0000 (16:12 -0800)]
[PATCH] C99 initialisers

C99 initialiser conversion; some from Rusty, some from me.

23 years ago[PATCH] add missing __exit specifications
David Howells [Wed, 6 Nov 2002 23:56:40 +0000 (15:56 -0800)]
[PATCH] add missing __exit specifications

This adds some missing __exit specifications which lead to a failure to
link the AFS code directly into the kernel.

23 years agoMerge bk://ppc.bkbits.net/for-linus-ppc
Linus Torvalds [Wed, 6 Nov 2002 23:55:02 +0000 (15:55 -0800)]
Merge bk://ppc.bkbits.net/for-linus-ppc
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoPPC32: Increase max kernel size in boot wrapper, fix compile warnings
Paul Mackerras [Fri, 8 Nov 2002 09:31:15 +0000 (20:31 +1100)]
PPC32: Increase max kernel size in boot wrapper, fix compile warnings

23 years agoPPC32: Fixes for the Makefiles under arch/ppc/boot.
Paul Mackerras [Fri, 8 Nov 2002 09:18:12 +0000 (20:18 +1100)]
PPC32: Fixes for the Makefiles under arch/ppc/boot.

With these changes the boot wrapper successfully builds,
although this may not be the absolute best way to do things.

23 years agoPPC32: Define CLONE_UNTRACED for assembler code, fix a too-long branch
Paul Mackerras [Fri, 8 Nov 2002 09:14:54 +0000 (20:14 +1100)]
PPC32: Define CLONE_UNTRACED for assembler code, fix a too-long branch

23 years agoPPC32: Remove AFLAGS for arch/ppc/mm/hashtable.o, not needed now.
Paul Mackerras [Fri, 8 Nov 2002 09:12:28 +0000 (20:12 +1100)]
PPC32: Remove AFLAGS for arch/ppc/mm/hashtable.o, not needed now.

23 years agoPPC32: Remove powermac SCSI boot disk discovery code.
Paul Mackerras [Fri, 8 Nov 2002 09:10:42 +0000 (20:10 +1100)]
PPC32: Remove powermac SCSI boot disk discovery code.

This didn't compile since sd_find_target is gone, and is to
move into userspace anyway.

23 years agoPPC32: Improved support for PReP platforms, forward-ported from 2.4.
Paul Mackerras [Fri, 8 Nov 2002 09:08:02 +0000 (20:08 +1100)]
PPC32: Improved support for PReP platforms, forward-ported from 2.4.

23 years agoPPC32: More sensible arrangement of the sections in vmlinux.lds.S.
Paul Mackerras [Fri, 8 Nov 2002 09:06:18 +0000 (20:06 +1100)]
PPC32: More sensible arrangement of the sections in vmlinux.lds.S.

This moves the sections which are read-only (e.g. exception table,
kallsyms data) to go before the read/write data section, and the
feature fixup section into the init data area.  It also adds the
initramfs section.

23 years agoPPC32: Remove more #ifdefs now that the config defines suitable
Tom Rini [Fri, 8 Nov 2002 09:03:11 +0000 (20:03 +1100)]
PPC32: Remove more #ifdefs now that the config defines suitable
defaults for the advanced kernel config options.

23 years agoPPC32: Fix up the arch-specific export list.
Paul Mackerras [Fri, 8 Nov 2002 08:58:43 +0000 (19:58 +1100)]
PPC32: Fix up the arch-specific export list.

We need __div64_32 exported, and flush_icache_page is now a noop
so it shouldn't be exported.

23 years agoPPC32: Define default settings for advanced config options.
Tom Rini [Fri, 8 Nov 2002 08:56:17 +0000 (19:56 +1100)]
PPC32: Define default settings for advanced config options.

This simplifies the C code by removing some #ifdefs.

23 years agoMerge samba.org:/home/paulus/kernel/linux-2.5
Paul Mackerras [Fri, 8 Nov 2002 07:22:55 +0000 (18:22 +1100)]
Merge samba.org:/home/paulus/kernel/linux-2.5
into samba.org:/home/paulus/kernel/for-linus-ppc

23 years agoPPC32: make the idle loop able to be platform-specific.
Paul Mackerras [Fri, 8 Nov 2002 07:21:16 +0000 (18:21 +1100)]
PPC32: make the idle loop able to be platform-specific.

23 years agoPPC32: add new syscalls: lookup_dcookie, epoll_*, remap_file_pages
Paul Mackerras [Fri, 8 Nov 2002 07:13:29 +0000 (18:13 +1100)]
PPC32: add new syscalls: lookup_dcookie, epoll_*, remap_file_pages

23 years agoPPC32: define MAP_POPULATE, MAP_NONBLOCK, POLLREMOVE
Paul Mackerras [Fri, 8 Nov 2002 07:08:58 +0000 (18:08 +1100)]
PPC32: define MAP_POPULATE, MAP_NONBLOCK, POLLREMOVE

23 years agoPPC32: Make flush_icache_page a no-op, do the flush in update_mmu_cache.
Paul Mackerras [Fri, 8 Nov 2002 03:40:39 +0000 (14:40 +1100)]
PPC32: Make flush_icache_page a no-op, do the flush in update_mmu_cache.

23 years agoMerge bk://source.mvista.com/linux-2.5-for-paulus
Paul Mackerras [Thu, 7 Nov 2002 20:19:49 +0000 (07:19 +1100)]
Merge bk://source.mvista.com/linux-2.5-for-paulus
into samba.org:/home/paulus/kernel/for-linus-ppc

23 years agoMerge master.kernel.org:/home/davem/BK/net-2.5
Linus Torvalds [Wed, 6 Nov 2002 23:54:07 +0000 (15:54 -0800)]
Merge master.kernel.org:/home/davem/BK/net-2.5
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoMerge nuts.ninka.net:/home/davem/src/BK/network-2.5
David S. Miller [Wed, 6 Nov 2002 17:47:18 +0000 (09:47 -0800)]
Merge nuts.ninka.net:/home/davem/src/BK/network-2.5
into nuts.ninka.net:/home/davem/src/BK/net-2.5

23 years agoMerge nuts.ninka.net:/home/davem/src/BK/sparcwork-2.5
David S. Miller [Wed, 6 Nov 2002 17:24:29 +0000 (09:24 -0800)]
Merge nuts.ninka.net:/home/davem/src/BK/sparcwork-2.5
into nuts.ninka.net:/home/davem/src/BK/sparc-2.5

23 years agodl2k net driver update from vendor:
Edward Peng [Wed, 6 Nov 2002 17:22:14 +0000 (12:22 -0500)]
dl2k net driver update from vendor:
* ethtool support
* changed default media to auto-negotiation
* fix disconnect bug
* fix RMON statistics overflow
* always use io mapping to access eeprom

23 years ago[BRIDGE]: Fix help docs.
Bart De Schuymer [Wed, 6 Nov 2002 17:16:09 +0000 (09:16 -0800)]
[BRIDGE]: Fix help docs.

23 years ago[CRYPTO]: minor updates
James Morris [Wed, 6 Nov 2002 16:38:49 +0000 (08:38 -0800)]
[CRYPTO]: minor updates
- Fixed min keysize bug for Blowfish (it is 32, not 64).
- Documentation updates.

23 years ago[CRYPTO]: Make sha256.c more palatable to GCCs optimizers.
David S. Miller [Wed, 6 Nov 2002 13:57:07 +0000 (05:57 -0800)]
[CRYPTO]: Make sha256.c more palatable to GCCs optimizers.

23 years ago[PATCH] Support latest NVRAM card from micromemory.
Neil Brown [Wed, 6 Nov 2002 13:08:14 +0000 (05:08 -0800)]
[PATCH] Support latest NVRAM card from micromemory.

Just a new PCI ID (and get twice the MegaHz :-).

23 years ago[PATCH] Make VT8653 work with AGP
Rob Weryk [Wed, 6 Nov 2002 13:08:07 +0000 (05:08 -0800)]
[PATCH] Make VT8653 work with AGP

This makes VT8653 (VIA Apollo Pro266T) work with AGP.  I had someone
test it and verify it works.

23 years ago[PATCH] Fix typo in sl82c105.c driver
Paul Mackerras [Wed, 6 Nov 2002 13:08:01 +0000 (05:08 -0800)]
[PATCH] Fix typo in sl82c105.c driver

This fixes a minor typo in sl82c105.c which stops it from compiling.

23 years ago[PATCH] Update powermac IDE driver
Paul Mackerras [Wed, 6 Nov 2002 13:07:55 +0000 (05:07 -0800)]
[PATCH] Update powermac IDE driver

This updates the powermac IDE driver in 2.5 so it uses the 2.5 kernel
interfaces and types rather than the 2.4 ones.  It also makes it use
blk_rq_map_sg rather than its own code to set up scatter/gather lists in
pmac_ide_build_sglist, and makes it use ide_lock instead of
io_request_lock.

23 years ago[PATCH] Update macserial driver
Paul Mackerras [Wed, 6 Nov 2002 13:07:50 +0000 (05:07 -0800)]
[PATCH] Update macserial driver

This updates the macserial driver in 2.5 so it compiles and works.  The
main changes are to use schedule_work instead of task queues and BHs.
The patch also removes the wait_key method.

I know we need to change macserial to use the new serial
infrastructure.  I'm posting this patch in case it is useful to anyone
trying to compile up a kernel for a powermac at the moment.

23 years ago[PATCH] epoll bits 0.34
Davide Libenzi [Wed, 6 Nov 2002 13:07:44 +0000 (05:07 -0800)]
[PATCH] epoll bits 0.34

 - Some constant adjusted
 - Comments added
 - Better hash initialization
 - Correct timeout setup
 - Added __KERNEL__ bypass to avoid userspace inclusion problems
 - Cleaned up locking
 - Function poll_init_wait() now calls poll_init_wait_ex()
 - Event return fix ( Jay Vosburgh )
 - Use <linux/hash.h> for the hash

23 years ago[PATCH] kNFSd: Make sure svc_process releases response even on error.
Neil Brown [Wed, 6 Nov 2002 13:00:53 +0000 (05:00 -0800)]
[PATCH] kNFSd: Make sure svc_process releases response even on error.

If a rpc operation indicates that response should
be dropped (e.g. kmalloc failure) we must still
call pc_release to release anything it may have allocated.

23 years ago[PATCH] kNFSd: Convert readlink to use a separate page for returning symlink contents.
Neil Brown [Wed, 6 Nov 2002 13:00:46 +0000 (05:00 -0800)]
[PATCH] kNFSd: Convert readlink to use a separate page for returning symlink contents.

This allows NFSv3 to manage 4096byte symlinks.

Also remove now-unused svcbuf_reserver function.

This was used to reserve space in output buffer
for 'data', but now this is stored in separate page.

23 years ago[PATCH] kNFSd: Make sure final xdr_buf.len is correct on server reply
Neil Brown [Wed, 6 Nov 2002 13:00:40 +0000 (05:00 -0800)]
[PATCH] kNFSd: Make sure final xdr_buf.len is correct on server reply

rq_res->len was not always updated properly.
It is only needed in the sendto routine, so we calculate
it just before that is called, and don't bother
updating it anywhere else.

23 years ago[PATCH] kNFSd: Support zero-copy read for NFSD
Neil Brown [Wed, 6 Nov 2002 13:00:35 +0000 (05:00 -0800)]
[PATCH] kNFSd: Support zero-copy read for NFSD

From Hirokazu Takahashi <taka@valinux.co.jp>

This patch changes read and readdir in nfsd.

read:
  If the file supports readpage, we use it to collect pages out of the page
  cache and to attache them directly to the outgoing nfs reply.
  The reduces the number of copies by one, and if the filesystem/device
  driver didn't copy the data, and if the network card can support not copying
  the data, then you get zero-copy reads.

readdir:
  A separate page is used for stoing the readdir response so that a fill
  PAGE_SIZE bytes of reply can be supported.

23 years ago[PATCH] kNFSd: Use ->sendpage to send nfsd (and lockd) replies.
Neil Brown [Wed, 6 Nov 2002 13:00:28 +0000 (05:00 -0800)]
[PATCH] kNFSd: Use ->sendpage to send nfsd (and lockd) replies.

From Hirokazu Takahashi <taka@valinux.co.jp>

As all rpc server replies are now in well defined pages,
we can use ->sendpage to send these replies, and so
make use for zero-copy transmit on network cards that
support it.

23 years ago[PATCH] md: Fix another two bug in raid5
Neil Brown [Wed, 6 Nov 2002 13:00:21 +0000 (05:00 -0800)]
[PATCH] md: Fix another two bug in raid5

 A partial block write over a block on a failed device
 would need to pre-read that block, which means pre-read all
 blocks in stripe and generate that block.  But the generate-block
 code never checked for this possibility, so it wouldn't happen.

23 years ago[PATCH] md: Fix bug in raid5
Neil Brown [Wed, 6 Nov 2002 13:00:15 +0000 (05:00 -0800)]
[PATCH] md: Fix bug in raid5

When analysing a stripe in handle_stripe we set bits
 R5_Wantread or R5_Wantwrite
to indicate if a read or write is needed.  We don't actually schedule the
IO immediately as this is done under a spinlock (sh->lock) and
generic_make_request can block.  Instead we check these bits after
the lock has been lifted and then schedule the IO.

But once the lock has been lifted we aren't safe against multiple
access, and it is possible that the IO will be scheduled never, or twice.

So, we use test_and_clear to check and potentially schedule the IO.

This wasn't a problem in 2.4 because the equivalent information was
stored on the stack instead of in the stripe.

We also make sure bi_io_vec[0] has correct values as a previous
call to generic_make_request may have changed them.

23 years ago[PATCH] md: Fix assorted raid1 problems.
Neil Brown [Wed, 6 Nov 2002 13:00:09 +0000 (05:00 -0800)]
[PATCH] md: Fix assorted raid1 problems.

From Angus Sawyer <angus.sawyer@dsl.pipex.com>:

1. Null pointer dereference in end_sync_read

   r1_bio->read_disk is not initialised correctly in sync_request .

   this is used in end_sync_read to reference the structure
   conf->mirror[read_disk].rdev which with one disk missing is NULL.

2. Null pointer dereference in mempool_free()

   This is a race between close_sync() conf->r1_bufpool =3D NULL and put_buf()
   mempool_free().

   bio completion -> resume_device -> put_buf -> mempool_free(r1_bufpool)
|
   [ wakeup]
|
   close_sync() -> r1_bufpool = NULL;

   The patch attached  reorders the mempool_free before the  barrier is released
   and merges resume_device() into put_buf(), (they are only used together).
   Otherwise I have kept the locking and wakeups identical to the existing code.
   (maybe this could be streamlined)

3.  BUG() at close_sync() if (waitqueue_active(&conf->wait_resume).

   This occurs with and without the patch for (2).

   I think this is a false BUG().  From what I understand of the device barrier
   code, there is nothing wrong with make_request() waiting on wait_resume when
   this test is made.  Therefore I have removed it (the wait_idle test is still
   correct).

4. raid1 tries to start a resync if there is only one working drive,
   which is pretty pointless, and noisy.  We notice that special case and
   avoid the resync.

23 years ago[PATCH] md: Misc little raid fixes
Neil Brown [Wed, 6 Nov 2002 13:00:03 +0000 (05:00 -0800)]
[PATCH] md: Misc little raid fixes

Roughly in order of patch:

1/ xor.h is never needed in md.c
2/ set sb_loaded when we 'sync' mddev to rdev as well as when we load
   sb into rdev from disk.
2/ due to lifetime changes, active count can be 2 when we stop array
3/ due to lifetime changes, we need to explicitly clear the ->pers when we
   stop an array
4/ autostart can only work for 0.90.0 superblocks.
   None others would be silly enough to store device numbers for
   all devices in the superblock...
5/ we had lost the setting of 'sb' when auto-starting an array.
6/ Code currently calls export_rdev(start_rdev) when
   IS_ERR(start_rdev), which causes an oops.
7/ /proc/mdstat contents error:
       code does not take into account that delayed
       resyncs can wait with curr_resync = 1 or 2.
8/ There is a premature "return NOTIFY_DONE", that possibly was in
   for debugging once...

23 years ago[PATCH] Fix typo in nfs_readpages.
Trond Myklebust [Wed, 6 Nov 2002 12:55:28 +0000 (04:55 -0800)]
[PATCH] Fix typo in nfs_readpages.

Make sure we drain the entire list of pages that failed to get added
to the mapping.

23 years agoMerge bk://linux-scsi.bkbits.net/scsi-dledford
Linus Torvalds [Wed, 6 Nov 2002 12:54:32 +0000 (04:54 -0800)]
Merge bk://linux-scsi.bkbits.net/scsi-dledford
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoMerge master.kernel.org:/home/hch/BK/xfs/linux-2.5
Linus Torvalds [Wed, 6 Nov 2002 12:53:39 +0000 (04:53 -0800)]
Merge master.kernel.org:/home/hch/BK/xfs/linux-2.5
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoexport find_trylock_page for XFS
Christoph Hellwig [Wed, 6 Nov 2002 23:39:06 +0000 (00:39 +0100)]
export find_trylock_page for XFS

23 years ago[XFS] Fix compilation with ACLs enabled
Christoph Hellwig [Wed, 6 Nov 2002 23:27:24 +0000 (00:27 +0100)]
[XFS] Fix compilation with ACLs enabled

SGI Modid: 2.5.x-xfs:slinx:132214a

23 years ago[XFS] Fix up some Kconfig merging issues
Christoph Hellwig [Wed, 6 Nov 2002 22:42:59 +0000 (23:42 +0100)]
[XFS] Fix up some Kconfig merging issues

23 years ago[XFS] Don't require ACL helpers for XFS
Christoph Hellwig [Wed, 6 Nov 2002 22:35:22 +0000 (23:35 +0100)]
[XFS] Don't require ACL helpers for XFS

SGI Modid: 2.5.x-xfs:slinx:132176a

23 years ago[XFS] Add XFS_POSIX_ACL to control ACL compilation in xfs
Stephen Lord [Wed, 6 Nov 2002 22:33:34 +0000 (23:33 +0100)]
[XFS] Add XFS_POSIX_ACL to control ACL compilation in xfs

SGI Modid: 2.5.x-xfs:slinx:132045a

23 years ago[XFS] remove excess vn_remove from the unmount path
Stephen Lord [Wed, 6 Nov 2002 22:24:48 +0000 (23:24 +0100)]
[XFS] remove excess vn_remove from the unmount path

SGI Modid: 2.5.x-xfs:slinx:131939a

23 years ago[XFS] remove VPURGE
Stephen Lord [Wed, 6 Nov 2002 22:22:28 +0000 (23:22 +0100)]
[XFS] remove VPURGE

SGI Modid: 2.5.x-xfs:slinx:131630a

23 years ago[XFS] Remove a couple other readonly device change remnants
Eric Sandeen [Wed, 6 Nov 2002 22:20:45 +0000 (23:20 +0100)]
[XFS] Remove a couple other readonly device change remnants

SGI Modid: 2.5.x-xfs:slinx:131565a

23 years ago[XFS] Be more careful about quota state changes on ro-devices
Eric Sandeen [Wed, 6 Nov 2002 22:18:46 +0000 (23:18 +0100)]
[XFS] Be more careful about quota state changes on ro-devices
We can't allow quota state changes on a read-only device,
this would kick of a failing transaction & shut down the fs.
Previously the test was quota/no quota but we need to disallow
any change wrt user and/or group quota state.

SGI Modid: 2.5.x-xfs:slinx:131554a

23 years ago[XFS] break out the allocator specific parts of the xfs I/O path into
Stephen Lord [Wed, 6 Nov 2002 22:17:04 +0000 (23:17 +0100)]
[XFS] break out the allocator specific parts of the xfs I/O path into
a separate file, xfs_iomap.c out of xfs_lrw.c. Remove some parts
of the code which were not doing anything for us. This is step
one in some major reorgs of this code.

SGI Modid: 2.5.x-xfs:slinx:131524a

23 years ago[XFS] narrow down comment
Russell Cattelan [Wed, 6 Nov 2002 22:15:09 +0000 (23:15 +0100)]
[XFS] narrow down comment

SGI Modid: 2.5.x-xfs:slinx:131504a

23 years ago[XFS] fix loop termination logic in xfs_sync
Stephen Lord [Wed, 6 Nov 2002 22:13:30 +0000 (23:13 +0100)]
[XFS] fix loop termination logic in xfs_sync

SGI Modid: 2.5.x-xfs:slinx:131490a

23 years ago[XFS] Fix fsx corruption.
Russell Cattelan [Wed, 6 Nov 2002 22:11:38 +0000 (23:11 +0100)]
[XFS] Fix fsx corruption.

SGI Modid: 2.5.x-xfs:slinx:131438a