- 29 Oct, 2004 15 commits
-
-
Mingming Cao authored
rbtree implementation and other changes From: Stephen Tweedie <sct@redhat.com> contributions From: Badari Pulavarty <pbadari@us.ibm.com> and probably me. This is the ext3 block reservation patch. It improves the layout of ext3 files by establishing, for each inode, reserved areas of the disk in which only that file can allocate blocks. Those reserved areas are managed in an rbtree, via the in-core inode. It's a bit like ext2 preallocation only stronger in that it can span already-allocated blocks, including the per-blockgroup inode tables and bitmaps. The patch fixes ext3's worst performance problem: disastrous layout when multiple files are being concurrently grown. It increases the size of the inode by rather a lot. A todo item is to dynamically allocate the `struct reserve_window_node', so we don't need to carry this storage for inodes which aren't opened for writing. The feature is enabled by mounting with the "reservation" mount option. Reservations default to "off". Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Mingming Cao authored
Cleans up the old ext3 preallocation code carried from ext2 but turned off. Signed-off-by:
-
Paolo \'Blaisorblade\' Giarrusso authored
While changing the TTY layer, an API parameter was removed, so it was removed by almost all calls, changing their prototype. But one use of one such function was not updated, breaking UML compilation. This is the fix. Signed-off-by:
Paolo 'Blaisorblade' Giarrusso <blaisorblade_spam@yahoo.it> Signed-off-by:
-
Hirokazu Takata authored
This patch fixes a typo of arch/m32r/lib/delay.c. It is required to fix a compile error for non-CONFIG_ISA_DUAL_ISSUE configuration. NOTE: The m32r has a kind of ISA (instruction set architecture) like a 2-way VLIW. Originally the m32r has two types of 16-bit/32-bit instructions. Only 16-bit instruction pair can be executed in parallel on a m32r chip with dual issuing support (e.g. M32700). According to the insturction dual issuing of the m32r, a programmer/compiler can explicitly put two 16-bit instructions into an instruction word, which are to be executed in parallel. ex. Assembly coding style of two 16-bit instruncions; insn A and B: 1) insn A insn B ; sequential execution (implicit) 2) insn A -> insn B ; sequential execution (explicit) 3) insn A || insn B ; parallel execution Signed-off-by:Hirokazu Takata <takata@linux-m32r.org> Signed-off-by:
-
Andi Kleen authored
Straightforward patch to remove a warning in genapic about a unused function. Independently done by Chris Wright too. Signed-off-by:
Andi Kleen <ak@muc.de> Signed-off-by:
-
Andi Kleen authored
genapic broke early safe_smp_processor_id(), especially when you got a WARN_ON or oops early it would loops forever in show_trace. The reason was that the x86_cpu_to_apicid array wasn't correctly initialized. This patch fixes this by just testing for this case. Orginally from James Cleverdon Acked-by: Andi Kleen <ak@muc.de> Signed-off-by:
-
Benjamin Herrenschmidt authored
Now that pci_get_legacy_ide_irq() support has been merged, it's time to enable use of it by the Maple platform code. Signed-off-by:
Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by:
-
Olof Johansson authored
Nathan Lynch pointed this out: The CPU sibling map is never initialized on iSeries. This makes the scheduler very unhappy if CONFIG_SCHED_SMT is enabled, causing an oops in find_busiest_group during boot. Below patch adds the expected init. Signed-off-by:
Olof Johansson <olof@austin.ibm.com> Signed-off-by:
-
Stephen Rothwell authored
The generic irq patches broke pci irqs on ppc64 iSeries. Signed-off-by:
Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by:
-
Benjamin Herrenschmidt authored
Tom's recent irq patch broke PowerMac (and possibly others). I think he forgot that PReP, CHRP and PowerMac are all built together in a single kernel image, thus all of those arch_initcall's will end up beeing called, even on the wrong machine... Signed-off-by:
-
Matt Porter authored
The patch fixes these warnings by doing two things: 1) Add the argument to the printk. 2) Rearrange the ifdef to eliminate the unused variable and function warnings. Signed-off-by:Scott Anderson <sanders@mvista.com> Signed-off-by:
Matt Porter <mporter@kernel.crashing.org> Signed-off-by:
-
Andrew Morton authored
Suppress page allocation warnings arising from radix_tree_node_alloc() allocations on behalf of swapper_space: these are expected under heavy load. Signed-off-by:
-
Andrew Morton authored
The data structures which are set up by key_init() are used by exec(). And we're using exec() super-early via the hotplug events from do_basic_setup():driver_init(). So call key_init() directly, prior to driver_init(). Cc: David S. Miller <davem@davemloft.net> Cc: David Howells <dhowells@redhat.com> Signed-off-by:
-
Pavel Machek authored
This adds typechecking to suspend types and powerdown types. This should solve at least part of suspend type confusion. There should be no code changes generated by this one. Acked-by:
Patrick Mochel <mochel@digitalimplant.org> Signed-off-by:
-
Toshihiro Iwamoto authored
It seems that O_DIRECT write sometimes leaks memory. Signed-off-by:
-
- 28 Oct, 2004 25 commits
-
-
Andi Kleen authored
The x86-64 genapic patch that was recently merged missed some definitions and doesn't compile at all. This fixes it by adding the missing defines for genapic Signed-off-by:
-
Tom Rini authored
D'oh. Something got cut-off there on Randy's end. Obviously correct fix follows. Signed-off-by:
Tom Rini <trini@kernel.crashing.org> Signed-off-by:
-
Olaf Hering authored
this sequence of commands will hang the final rmmod. init=/bin/bash --login mount proc modprove -v ohci1394 rmmod ohci1394 modprobe -v ohci1394 rmmod ohci1394 -> hangs knodemgrd_0 exits on the first rmmod, but leaves nodemgr_serialize in down state. This fixes it for me. Signed-off-by:
Olaf Hering <olh@suse.de> Acked-by:
Ben Collins <bcollins@debian.org> Signed-off-by:
-
http://lia64.bkbits.net/linux-ia64-release-2.6.10Linus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
Stephen Rothwell authored
This merges two files (iSeries_IoMmTable.[ch]) into iSeries_pci.c. This allowed quite a few more things to become declared static. It then does some fairly mechanical cleanups in iSeries_pci.c (replacing studly caps, removing the last of the PCIFR() macros and removing a couple of empty or unused routines). There are no semantic changes. Signed-off-by:
-
Martin Wilck authored
The "Synchronize Arbitration IDs with APIC IDs" IPI does not make sense on Pentium 4 and Xeon because they don't have an APIC bus where arbitration IDs would be needed. The Intel Developers manual (Vol. 3, Ch. 8.6.1) says this IPI is unsupported on P4/Xeon. Using it may, according to Intel, lead to unpredictable results. Signed-off-by:
Martin Wilck <martin.wilck@fujitsu-siemens.com> Signed-off-by:
-
Stelian Pop authored
Signed-off-by:
Stelian Pop <stelian@popies.net> Signed-off-by:
-
Stelian Pop authored
Signed-off-by:
-
Stelian Pop authored
Leave the choice to the user. Signed-off-by:
-
Stelian Pop authored
Signed-off-by:
-
Stelian Pop authored
* feed most of special keys through the input subsystem * initialize two separate input devices: a mouse like one for the jogdial and a keyboard like one for the special keys * add support for SONYPI_EVENT_FNKEY_RELEASED Many people participated in a way or another to this patch, including Daniel K. <daniel@cluded.net>, Bastien Nocera <hadess@hadess.net>, Dmitry Torokhov <dtor@mail.ru> and Vojtech Pavlik <vojtech@suse.cz>. Signed-off-by:
-
Stelian Pop authored
* switch from a sysdev to a platform device * drop old style PM code * use pci_get_device()/pci_dev_put() instead of pci_find_device() Patch originaly from Dmitry Torokhov <dtor@mail.ru>. Signed-off-by:
-
Stelian Pop authored
Signed-off-by:
-
Stelian Pop authored
* use module_param() instead of MODULE_PARM() and __setup() * use MODULE_VERSION() Signed-off-by:
-
Chris Wright authored
Early calls to userspace can invoke an execve() before any binfmt handlers are registered. Properly return an error in this case rather than 0. On at least one arch (x86_64) without this patch, the system will double fault on early attempts to call_usermodehelper. Signed-off-by:
Chris Wright <chrisw@osdl.org> Signed-off-by:
-
Chris Wedgwood authored
Remove build warning from drivers/char/random.c on 32-bit platforms. Signed-off-by:
Chris Wedgwood <cw@f00f.org> Signed-off-by:
-
Olaf Hering authored
__handle_sysrq already prints a newline, so the action_msg string doesnt need yet another newline. Signed-off-by:
-
Gerd Knorr authored
- Make bttv check the device_register return value. - Yet another unchecked return value fixed. Signed-off-by:
Gerd Knorr <kraxel@bytesex.org> Signed-off-by:
-
Jens Axboe authored
There are two issues with online io scheduler switching that this patch addresses. The first is pretty simple - it concerns racing with scheduler removal on switch. elevator_find() does not grab a reference to the io scheduler, so before elevator_attach() is run it could go away. Add elevator_get() to solve that. Second issue is the flushing out of requests that is needed before switching can be problematic with requests that aren't allocated in the block layer (such as requests on the stack of a process). The problem is that we don't know when they will have finished, and most io schedulers need to access the elevator structures on io completion. This can be fixed by adding an intermedia step that switches to noop, since it doesn't need to touch anything but the request_queue. The queue drain can then safely be split into two operations - one that waits for file system requests, and one that waits for the queue to completely empty. Requests arriving after the first drain will get stuck in a seperate queue list. Signed-off-by:
Jens Axboe <axboe@suse.de> Signed-off-by:
-
Manfred Spraul authored
register_sysctl_table() fails if sysctl support is not compiled into the kernel. The POSIX message queue subsystem aborted it's initialization if register_sysctl_table() fails, and that causes an oops in sys_mq_open(). The patch fixes that by ignoring failures from register_sysctl_table(). Signed-off-by; Manfred Spraul <manfred@colorfullife.com> Signed-off-by:
Adrian Bunk <bunk@stusta.de> Signed-off-by:
-
Jens Axboe authored
Fix online switching issue with cfq v2. It does deferred clearing of e->elevator_data, which screws up the current io scheduler. It does not have to, so just remove it. Signed-off-by:
-
William Lee Irwin III authored
These two variables are long, long dead. This patch removes them. Signed-off-by:
William Irwin <wli@holomorphy.com> Signed-off-by:
-
Colin Leroy authored
This patch fixes via-pmu.c compilation when CONFIG_PMAC_PBOOK is not set. Looks like the #endif was one line too late. Signed-off-by:
Colin Leroy <colin@colino.net> Acked-by:
-
Christoph Hellwig authored
It's been totally obsoleted by the input layer Acked-by:
Alan Cox <alan@redhat.com> Signed-off-by:
-
Christoph Hellwig authored
This was used by the early irqstacks implementation on s390 and has been replaced by __ARCH_HAS_DO_SOFTIRQ now. Signed-off-by:
-
