Jan Engelhardt [Wed, 17 Oct 2007 06:30:15 +0000 (23:30 -0700)]
menuconfig: transform NLS and DLM menus
Changes NLS and DLM menus into a 'menuconfig' object so that it can be
disabled at once without having to enter the menu first to disable the config
option.
Rusty Russell [Wed, 17 Oct 2007 06:30:13 +0000 (23:30 -0700)]
Delay creation of khcvd thread
This changes hvc_init() to be called only when someone actually uses the
hvc_console driver. Dave Jones complained when profiling bootup.
hvc_console used to only be for Power aka pSeries: now lguest and Xen both
want it built-in in case the kernel is a guest under one of those, even
though usually it will be a native boot.
Andrey Mirkin [Wed, 17 Oct 2007 06:30:13 +0000 (23:30 -0700)]
change inotifyfs magic as the same magic is used for futexfs
Right now futexfs and inotifyfs have one magic 0xBAD1DEA, that looks a
little bit confusing. Use 0xBAD1DEA as magic for futexfs and 0x2BAD1DEA as
magic for inotifyfs.
Olaf Hering [Wed, 17 Oct 2007 06:30:12 +0000 (23:30 -0700)]
increase AT_VECTOR_SIZE to terminate saved_auxv properly
include/asm-powerpc/elf.h has 6 entries in ARCH_DLINFO. fs/binfmt_elf.c
has 14 unconditional NEW_AUX_ENT entries and 2 conditional NEW_AUX_ENT
entries. So in the worst case, saved_auxv does not get an AT_NULL entry at
the end.
The saved_auxv array must be terminated with an AT_NULL entry. Make the
size of mm_struct->saved_auxv arch dependend, based on the number of
ARCH_DLINFO entries.
Cyrill Gorcunov [Wed, 17 Oct 2007 06:30:08 +0000 (23:30 -0700)]
UDF: coding style fixups
This patch does additional coding style fixup. Initially the code is being
distorted by Lindent (in my patches sent not very long ago) and fixed in
the followup patches but this stuff was accidently missed.
New and old compiled files were compared with cmp to check for being
identically. So the patch will not break the kernel.
Alan Cox [Wed, 17 Oct 2007 06:30:07 +0000 (23:30 -0700)]
tty: expose new methods needed for drivers to get termios right
This adds three new functions (or in one case to be more exact makes it
always available)
tty_termios_copy_hw
Copies all the hardware settings from one termios structure to the other.
This is intended for drivers that support little or no hardware setting
tty_termios_encode_baud_rate
Allows you to set the input and output baud rate in a termios structure. A
driver is supposed to set the resulting baud rate from a request so most
will want to use this function to set the resulting input and output rates
to match the hardware values. Internally it knows about keeping Bxxx
encoding when possible to maximise compatibility.
tty_encode_baud_rate
As above but for the tty's own current termios structure
I suspect this will initially need some tweaking as it gets enabled by
driver patches over the next few mm cycles so consider this lot -mm only
for the moment so it can stabilize and end up neat before it goes to base.
I've tried not to break any obscure architectures - if you get a speed you
can't represent the code will print warnings on non updated termios systems
but not break.
Once this is merged and seems sane I've got a growing pile of driver
updates to use it - notably for USB serial drivers.
Alan Cox [Wed, 17 Oct 2007 06:30:06 +0000 (23:30 -0700)]
ide-cd is unmaintained
I simply don't have any old IDE systems any more or time to really look
after this. Nobody responded to the previous linux-ide mail about
maintainers so...
Borislav Petkov [Wed, 17 Oct 2007 06:30:05 +0000 (23:30 -0700)]
fs/isofs/namei.c: Remove uninitialized local vars warning
shut up those:
fs/isofs/namei.c: In function 'isofs_lookup':
fs/isofs/namei.c:161: warning: 'offset' may be used uninitialized in this function
fs/isofs/namei.c:161: warning: 'block' may be used uninitialized in this function
By the way, they get overwritten at the end of isofs_find_entry().
Lepton Wu [Wed, 17 Oct 2007 06:30:04 +0000 (23:30 -0700)]
reiserfs: workaround for dead loop in finish_unfinished
There is possible dead loop in finish_unfinished function.
In most situation, the call chain iput -> ... -> reiserfs_delete_inode ->
remove_save_link will success. But for some reason such as data
corruption, reiserfs_delete_inode fails on reiserfs_do_truncate ->
search_for_position_by_key.
Then remove_save_link won't be called. We always get the same
"save_link_key" in the while loop in finish_unfinished function. The
following patch adds a check for the possible dead loop and just remove
save link when deap loop.
Denis V. Lunev [Wed, 17 Oct 2007 06:29:53 +0000 (23:29 -0700)]
shrink_dcache_sb speedup
This patch makes shrink_dcache_sb consistent with dentry pruning policy.
On the first pass we iterate over dentry unused list and prepare some
dentries for removal.
However, since the existing code moves evicted dentries to the beginning of
the LRU it can happen that fresh dentries from other superblocks will be
inserted *before* our dentries.
This can result in significant slowdown of shrink_dcache_sb(). Moreover,
for virtual filesystems like unionfs which can call dput() during dentries
kill existing code results in O(n^2) complexity.
We observed 2 minutes shrink_dcache_sb() with only 35000 dentries.
To avoid this effects we propose to isolate sb dentries at the end
of LRU list.
Remove the old-fashioned lk201 driver under drivers/tc/ that used to be
used by the old dz.c and zs.c drivers, which is now orphan code referred to
from nowhere and does not build anymore. A modern replacement is available
as drivers/input/keyboard/lkkbd.c.
There are no plans to do anything about this piece of code and it does not
fit anywhere anymore, so it is not just a matter of maintenance or the lack
of. There are still some bits that might be added to the new lkkbd.c
driver based on the old code, and the embedded hardware documentation which
is otherwise quite hard to get hold of might be useful to keep too. Both
of these can be done separately though. RIP.
Lepton Wu [Wed, 17 Oct 2007 06:29:50 +0000 (23:29 -0700)]
reiserfs: fix kernel panic on corrupted directory
When reading corrupted reiserfs directory data, d_reclen could be a
negative number or a big positive number, this can lead to kernel panic or
oop. The following patch adds a sanity check.
David Howells [Wed, 17 Oct 2007 06:29:46 +0000 (23:29 -0700)]
KEYS: Make request_key() and co fundamentally asynchronous
Make request_key() and co fundamentally asynchronous to make it easier for
NFS to make use of them. There are now accessor functions that do
asynchronous constructions, a wait function to wait for construction to
complete, and a completion function for the key type to indicate completion
of construction.
Note that the construction queue is now gone. Instead, keys under
construction are linked in to the appropriate keyring in advance, and that
anyone encountering one must wait for it to be complete before they can use
it. This is done automatically for userspace.
The following auxiliary changes are also made:
(1) Key type implementation stuff is split from linux/key.h into
linux/key-type.h.
(2) AF_RXRPC provides a way to allocate null rxrpc-type keys so that AFS does
not need to call key_instantiate_and_link() directly.
(3) Adjust the debugging macros so that they're -Wformat checked even if
they are disabled, and make it so they can be enabled simply by defining
__KDEBUG to be consistent with other code of mine.
Chris Mason [Wed, 17 Oct 2007 06:29:44 +0000 (23:29 -0700)]
try to reap reiserfs pages left around by invalidatepage
reiserfs_invalidatepage will refuse to free pages if they have been logged
in data=journal mode, or were pinned down by a data=ordered operation. For
data=journal, this is fairly easy to trigger just with fsx-linux, and it
results in a large number of pages hanging around on the LRUs with
page->mapping == NULL.
Calling try_to_free_buffers when reiserfs decides it is done with the page
allows it to be freed earlier, and with much less VM thrashing. Lock
ordering rules mean that reiserfs can't call lock_page when it is releasing
the buffers, so TestSetPageLocked is used instead. Contention on these
pages should be rare, so it should be sufficient most of the time.
Randy Dunlap [Wed, 17 Oct 2007 06:29:44 +0000 (23:29 -0700)]
maintainers: linux-omap list is subscribers only
You are not allowed to post to this mailing list, and your message has been
automatically rejected. If you think that your messages are being rejected
in error, contact the mailing list owner at [email protected].
Ralf Baechle [Wed, 17 Oct 2007 06:29:42 +0000 (23:29 -0700)]
Remove dma_cache_(wback|inv|wback_inv) functions
dma_cache_(wback|inv|wback_inv) were the earliest attempt on a generalized
cache managment API for I/O purposes. Originally it was basically the raw
MIPS low level cache API exported to the entire world. The API has
suffered from a lack of documentation, was not very widely used unlike it's
more modern brothers and can easily be replaced by dma_cache_sync. So
remove it rsp. turn the surviving bits back into an arch private API, as
discussed on linux-arch.
Satyam Sharma [Wed, 17 Oct 2007 06:29:39 +0000 (23:29 -0700)]
I2O: Fix "defined but not used" build warnings
drivers/message/i2o/exec-osm.c:539: warning: `i2o_exec_lct_notify' defined but not used
comes when CONFIG_I2O_LCT_NOTIFY_ON_CHANGES=n, because its only callsite
is #ifdef'ed as such. So let's #ifdef the function definition also. Also
move the definition to before the callsite, to get rid of forward prototype.
Andy Whitcroft [Wed, 17 Oct 2007 06:29:38 +0000 (23:29 -0700)]
update checkpatch.pl to version 0.10
This version brings a number of new checks, and a number of bug
fixes. Of note:
- better categorisation and space checks for dual use unary/binary
operators
- warn on deprecated use of {SPIN,RW}_LOCK_UNLOCKED
- check if/for/while with trailing ';' for hanging statements
- detect DOS line endings
- detect redundant casts for kalloc()
Andy Whitcroft (18):
Version: 0.10
asmlinkage is also a storage type
pull out inline specifiers
allow only some operators before a unary operator
parenthesised values may span line ends
add additional attribute matching
handle sparse annotations within pointer type space checks
support alternative function definition syntax for typedefs
check if/for/while with trailing ';' for hanging statements
fix output format for case checks
deprecate SPIN_LOCK_UNLOCKED and RW_LOCK_UNLOCKED
allow complex macros with bracketing braces
detect and report DOS line endings
fastcall is a valid function attribute
bracket spacing is ok for 'for'
categorise operators into unary/binary/definitions
add heuristic to pick up on unannotated types
remove spurious warnings from cat_vet
Dave Jones (1):
Make checkpatch warn about pointless casting of kalloc returns.
Bill Nottingham [Wed, 17 Oct 2007 06:29:38 +0000 (23:29 -0700)]
add CONFIG_VT_UNICODE
As of now, the kernel defaults to non-unicode and XLATE for the keyboard.
We've been changing this in Fedora, but that requires patching the defaults
in the kernel.
The attached introduces CONFIG_VT_UNICODE, which sets the console in
unicode mode by default on boot, including both the virtual terminal and
the keyboard driver.
Peter Lund [Wed, 17 Oct 2007 06:29:35 +0000 (23:29 -0700)]
avoid negative (and full-width) shifts in radix-tree.c
Negative shifts are not allowed in C (the result is undefined). Same thing
with full-width shifts.
It works on most platforms but not on the VAX with gcc 4.0.1 (it results in an
"operand reserved" fault).
Shifting by more than the width of the value on the left is also not
allowed. I think the extra '>> 1' tacked on at the end in the original
code was an attempt to work around that. Getting rid of that is an extra
feature of this patch.
Here's the chapter and verse, taken from the final draft of the C99
standard ("6.5.7 Bitwise shift operators", paragraph 3):
"The integer promotions are performed on each of the operands. The
type of the result is that of the promoted left operand. If the
value of the right operand is negative or is greater than or equal
to the width of the promoted left operand, the behavior is
undefined."
Thank you to Jan-Benedict Glaw, Christoph Hellwig, Maciej Rozycki, Pekka
Enberg, Andreas Schwab, and Christoph Lameter for review. Special thanks
to Andreas for spotting that my fix only removed half the undefined
behaviour.
Jan Beulich [Wed, 17 Oct 2007 06:29:34 +0000 (23:29 -0700)]
store __setup_str_* in a more compact way
__setup_str_* are referenced only during boot, hence there's no need to
waste image space for aligning these strings (with the aim of improving
performance).
Jan Beulich [Wed, 17 Oct 2007 06:29:33 +0000 (23:29 -0700)]
handle recursive calls to bust_spinlocks()
Various architectures may call bust_spinlocks() recursively; the function
itself, however, doesn't appear to be meant to be called in this manner.
Nevertheless, this doesn't appear to be a problem as long as
bust_spinlocks(0) doesn't get called twice in a row (otherwise,
unblank_screen() may enter the scheduler). However, at least on i386 die()
has been capable of returning (and on other architectures this should
really be that way, too) when notify_die() returns NOTIFY_STOP.
Short of getting a reply to a respective query, this patch makes
bust_spinlocks() increment/decrement oops_in_progress, and wake klogd only
when the count drops back to zero.
Robert P. J. Day [Wed, 17 Oct 2007 06:29:32 +0000 (23:29 -0700)]
Add a "rounddown_pow_of_two" routine to log2.h
To go along with the existing "roundup_pow_of_two" routine, add one for
rounding down since that operation appears to crop up on a regular basis in
the source tree.
Jan Kara [Wed, 17 Oct 2007 06:29:31 +0000 (23:29 -0700)]
quota: send messages via netlink
Implement sending of quota messages via netlink interface. The advantage
is that in userspace we can better decide what to do with the message - for
example display a dialogue in your X session or just write the message to
the console. As a bonus, we can get rid of problems with console locking
deep inside filesystem code once we remove the old printing mechanism.
Grant Grundler was asking for more detail about correct usage of local
atomic operations and suggested adding the resulting summary to
local_ops.txt.
"Please add a bit more detail. If DaveM is correct (he normally is), then
there must be limits on how the local_t can be used in the kernel process
and interrupt contexts. I'd like those rules spelled out very clearly
since it's easy to get wrong and tracking down such a bug is quite
painful."
Robert P. J. Day [Wed, 17 Oct 2007 06:29:27 +0000 (23:29 -0700)]
Remove valueless definition of hard-selected RAMFS option
Since CONFIG_RAMFS is currently hard-selected to "y", and since
Documentation/filesystems/ramfs-rootfs-initramfs.txt reads as follows:
"The amount of code required to implement ramfs is tiny, because all the
work is done by the existing Linux caching infrastructure. Basically,
you're mounting the disk cache as a filesystem. Because of this, ramfs is
not an optional component removable via menuconfig, since there would be
negligible space savings."
It seems pointless to leave this as a Kconfig entry.
Steve Cameron [Wed, 17 Oct 2007 06:27:37 +0000 (23:27 -0700)]
cciss: fix error reporting for SG_IO
This fixes a problem with the way cciss was filling out the "errors" field
of the request structure upon completion of requests. Previously, it just
put a 1 or a 0 in there and used the negation of this as the uptodate
parameter to one of the functions in the block layer, being a block device.
For the SG_IO ioctl, this was not sufficient, and we noticed that, for
example, sg_turs from sg3_utils did not correctly detect problems due to
cciss having set rq->errors incorrectly.
Paul Clements [Wed, 17 Oct 2007 06:27:37 +0000 (23:27 -0700)]
NBD: allow hung network I/O to be cancelled
Allow NBD I/O to be cancelled when a network outage occurs. Previously, I/O
would just hang, and if enough I/O was hung in nbd, the system (at least
user-level) would completely hang until a TCP timeout (default, 15 minutes)
occurred.
The patch introduces a new ioctl NBD_SET_TIMEOUT that allows a transmit
timeout value (in seconds) to be specified. Any network send that exceeds the
timeout will be cancelled and the nbd connection will be shut down. I've
tested with various timeout values and 6 seconds seems to be a good choice for
the timeout. If the NBD_SET_TIMEOUT ioctl is not called, you get the old (I/O
hang) behavior.
Paul Clements [Wed, 17 Oct 2007 06:27:36 +0000 (23:27 -0700)]
NBD: set uninitialized devices to size 0
This fixes errors with utilities (such as LVM's vgscan) that try to scan all
devices. Previously this would generate read errors when uninitialized nbd
devices were scanned:
# vgscan
Reading all physical volumes. This may take a while...
/dev/nbd0: read failed after 0 of 1024 at 0: Input/output error
/dev/nbd0: read failed after 0 of 1024 at 509804544: Input/output error
/dev/nbd0: read failed after 0 of 2048 at 0: Input/output error
/dev/nbd1: read failed after 0 of 1024 at 509804544: Input/output error
/dev/nbd1: read failed after 0 of 2048 at 0: Input/output error
From now on, uninitialized nbd devices will have size zero, which
prevents these errors.
Alan Cox [Wed, 17 Oct 2007 06:27:33 +0000 (23:27 -0700)]
CodingStyle: relax the 80-cole rule
I would suggest this change to make CodingStyle properly reflect the style
used by the kernel, rather than the current wording which is wishful
thinking and misleading, and comes from the same school of thought that
gets off on prescriptive grammar, latin and comp.std.c
Jan Beulich [Wed, 17 Oct 2007 06:27:32 +0000 (23:27 -0700)]
floppy: tolerate DMA channel unavailability
The floppy driver is already written to be able to operate in virtual DMA
mode. Thus it can easily be adjusted to tolerate failure from
fd_request_dma() as long as virtual DMA mode is not disallowed.
Avi Kivity [Wed, 17 Oct 2007 06:27:31 +0000 (23:27 -0700)]
Move PREEMPT_NOTIFIERS into an always-included Kconfig
Kconfig.preempt is not included on some archs (for example, m68k). On those
archs, the Kconfig machinery complains that KVM selects an undefined symbol
PREEMPT_NOTIFIERS (which lives in Kconfig.preempt).
So move the offending symbol into a Kconfig file which is included by
everyone.
Ken'ichi Ohmichi [Wed, 17 Oct 2007 06:27:30 +0000 (23:27 -0700)]
add-vmcore: add a prefix "VMCOREINFO_" to the vmcoreinfo macros
Add a prefix "VMCOREINFO_" to the vmcoreinfo macros. Old vmcoreinfo macros
were defined as generic names SYMBOL/SIZE/OFFSET /LENGTH/CONFIG, and it is
impossible to grep for them. So these names should be changed. This
discussion is the following:
http://www.ussg.iu.edu/hypermail/linux/kernel/0709.1/0415.html
Ken'ichi Ohmichi [Wed, 17 Oct 2007 06:27:28 +0000 (23:27 -0700)]
add-vmcore: add nodemask_t's size and NR_FREE_PAGES's value to vmcoreinfo_data
[2/3] Add nodemask_t's size and NR_FREE_PAGES's value to vmcoreinfo_data.
The dump filetering command 'makedumpfile'(v1.1.6 or before) had assumed
the above values, and it was not good from the reliability viewpoint.
So makedumpfile v1.2.0 came to need these values and I created the patch
to let the kernel output them.
makedumpfile site:
https://sourceforge.net/projects/makedumpfile/
Ken'ichi Ohmichi [Wed, 17 Oct 2007 06:27:28 +0000 (23:27 -0700)]
add-vmcore: cleanup the coding style according to Andrew's comments
[1/3] Cleanup the coding style according to Andrew's comments:
http://lists.infradead.org/pipermail/kexec/2007-August/000522.html
- vmcoreinfo_append_str() should have suitable __attribute__s so that
the compiler can check its use.
- vmcoreinfo_max_size should have size_t.
- Use get_seconds() instead of xtime.tv_sec.
- Use init_uts_ns.name.release instead of UTS_RELEASE.
Ken'ichi Ohmichi [Wed, 17 Oct 2007 06:27:27 +0000 (23:27 -0700)]
Add vmcoreinfo
This patch set frees the restriction that makedumpfile users should install a
vmlinux file (including the debugging information) into each system.
makedumpfile command is the dump filtering feature for kdump. It creates a
small dumpfile by filtering unnecessary pages for the analysis. To
distinguish unnecessary pages, it needs a vmlinux file including the debugging
information. These days, the debugging package becomes a huge file, and it is
hard to install it into each system.
To solve the problem, kdump developers discussed it at lkml and kexec-ml. As
the result, we reached the conclusion that necessary information for dump
filtering (called "vmcoreinfo") should be embedded into the first kernel file
and it should be accessed through /proc/vmcore during the second kernel.
(http://www.uwsg.iu.edu/hypermail/linux/kernel/0707.0/1806.html)
Dan Aloni created the patch set for the above implementation.
(http://www.uwsg.iu.edu/hypermail/linux/kernel/0707.1/1053.html)
And I updated it for multi architectures and memory models.
(http://lists.infradead.org/pipermail/kexec/2007-August/000479.html)
Andrew Morton [Wed, 17 Oct 2007 06:27:26 +0000 (23:27 -0700)]
binfmt_flat: warning fixes
Fix this lot:
fs/binfmt_flat.c: In function `decompress_exec':
fs/binfmt_flat.c:293: warning: label `out' defined but not used
fs/binfmt_flat.c: In function `load_flat_file':
fs/binfmt_flat.c:462: warning: unsigned int format, long int arg (arg 3)
fs/binfmt_flat.c:462: warning: unsigned int format, long int arg (arg 4)
fs/binfmt_flat.c:518: warning: comparison of distinct pointer types lacks a cast
fs/binfmt_flat.c:549: warning: passing arg 1 of `ksize' makes pointer from integer without a cast
fs/binfmt_flat.c:601: warning: passing arg 1 of `ksize' makes pointer from integer without a cast
fs/binfmt_flat.c: In function `load_flat_binary':
fs/binfmt_flat.c:116: warning: 'dummy' might be used uninitialized in this function
Oleg Nesterov [Wed, 17 Oct 2007 06:27:24 +0000 (23:27 -0700)]
do_sigaction: don't worry about signal_pending()
do_sigaction() returns -ERESTARTNOINTR if signal_pending(). The comment says:
* If there might be a fatal signal pending on multiple
* threads, make sure we take it before changing the action.
I think this is not needed. We should only worry about SIGNAL_GROUP_EXIT case,
bit it implies a pending SIGKILL which can't be cleared by do_sigaction.
Oleg Nesterov [Wed, 17 Oct 2007 06:27:23 +0000 (23:27 -0700)]
exec: RT sub-thread can livelock and monopolize CPU on exec
de_thread() yields waiting for ->group_leader to be a zombie. This deadlocks
if an rt-prio execer shares the same cpu with ->group_leader. Change the code
to use ->group_exit_task/notify_count mechanics.
This patch certainly uglifies the code, perhaps someone can suggest something
better.
Oleg Nesterov [Wed, 17 Oct 2007 06:27:23 +0000 (23:27 -0700)]
exec: consolidate 2 fast-paths
Now that we don't pre-allocate the new ->sighand, we can kill the first fast
path, it doesn't make sense any longer. At best, it can save one "list_empty()"
check but leads to the code duplication.
Oleg Nesterov [Wed, 17 Oct 2007 06:27:22 +0000 (23:27 -0700)]
exec: simplify the new ->sighand allocation
de_thread() pre-allocates newsighand to make sure that exec() can't fail after
killing all sub-threads. Imho, this buys nothing, but complicates the code:
- this is (mostly) needed to handle CLONE_SIGHAND without CLONE_THREAD
tasks, this is very unlikely (if ever used) case
- unless we already have some serious problems, GFP_KERNEL allocation
should not fail
- ENOMEM still can happen after de_thread(), ->sighand is not the last
object we have to allocate
Change the code to allocate the new ->sighand on demand.
Oleg Nesterov [Wed, 17 Oct 2007 06:27:22 +0000 (23:27 -0700)]
exec: simplify ->sighand switching
There is no any reason to do recalc_sigpending() after changing ->sighand.
To begin with, recalc_sigpending() does not take ->sighand into account.
This means we don't need to take newsighand->siglock while changing sighands.
rcu_assign_pointer() provides a necessary barrier, and if another process
reads the new ->sighand it should either take tasklist_lock or it should use
lock_task_sighand() which has a corresponding smp_read_barrier_depends().
Jeff Moyer [Wed, 17 Oct 2007 06:27:20 +0000 (23:27 -0700)]
aio: account I/O wait time properly
Some months back I proposed changing the schedule() call in
read_events to an io_schedule():
http://osdir.com/ml/linux.kernel.aio.general/2006-10/msg00024.html
This was rejected as there are AIO operations that do not initiate
disk I/O. I've had another look at the problem, and the only AIO
operation that will not initiate disk I/O is IOCB_CMD_NOOP. However,
this command isn't even wired up!
Given that it doesn't work, and hasn't for *years*, I'm going to
suggest again that we do proper I/O accounting when using AIO.
Paul E. McKenney [Wed, 17 Oct 2007 06:27:19 +0000 (23:27 -0700)]
Make rcutorture RNG use temporal entropy
Repost of http://lkml.org/lkml/2007/8/10/472 made available by request.
The locking used by get_random_bytes() can conflict with the
preempt_disable() and synchronize_sched() form of RCU. This patch changes
rcutorture's RNG to gather entropy from the new cpu_clock() interface
(relying on interrupts, preemption, daemons, and rcutorture's reader
thread's rock-bottom scheduling priority to provide useful entropy), and
also adds and EXPORT_SYMBOL_GPL() to make that interface available to GPLed
kernel modules such as rcutorture.
john stultz [Wed, 17 Oct 2007 06:27:18 +0000 (23:27 -0700)]
Use num_possible_cpus() instead of NR_CPUS for timer distribution
To avoid lock contention, we distribute the sched_timer calls across the
cpus so they do not trigger at the same instant. However, I used NR_CPUS,
which can cause needless grouping on small smp systems depending on your
kernel config. This patch converts to using num_possible_cpus() so we
spread it as evenly as possible on every machine.
Briefly tested w/ NR_CPUS=255 and verified reduced contention.
Chris Wright [Wed, 17 Oct 2007 06:27:18 +0000 (23:27 -0700)]
Use ERESTART_RESTARTBLOCK if poll() is interrupted by a signal
Lomesh reported poll returning EINTR during suspend/resume cycle. This is
caused by the STOP/CONT cycle that the freezer uses, generating a pending
signal for what in effect is an ignored signal. In general poll is a
little eager in returning EINTR, when it could try not bother userspace and
simply restart the syscall. Both select and ppoll do use ERESTARTNOHAND to
restart the syscall. Oleg points out that simply using ERESTARTNOHAND will
cause poll to restart with original timeout value. which could ultimately
lead to process never returning to userspace. Instead use
ERESTART_RESTARTBLOCK, and restart poll with updated timeout value.
Inspired by Manfred's use ERESTARTNOHAND in poll patch.
Adrian Bunk [Wed, 17 Oct 2007 06:27:17 +0000 (23:27 -0700)]
allow disabling DNOTIFY without EMBEDDED
Allow disabling DNOTIFY with CONFIG_EMBEDDED=n.
I'm currently running a kernel with dnotify disabled and I haven't run into
any problem. Is there any popular application left that breaks without
dnotify support in the kernel?
Note that this patch does not remove dnotify support, it still defaults to
"y", and the help text recommends enabling it.
Eric Sandeen [Wed, 17 Oct 2007 06:27:15 +0000 (23:27 -0700)]
limit minixfs printks on corrupted dir i_size
This attempts to address CVE-2006-6058
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2006-6058
first reported at http://projects.info-pull.com/mokb/MOKB-17-11-2006.html
Essentially a corrupted minix dir inode reporting a very large
i_size will loop for a very long time in minix_readdir, minix_find_entry,
etc, because on EIO they just move on to try the next page. This is
under the BKL, printk-storming as well. This can lock up the machine
for a very long time. Simply ratelimiting the printks gets things back
under control. Make the message a bit more informative while we're here.
Alexey Dobriyan [Wed, 17 Oct 2007 06:27:14 +0000 (23:27 -0700)]
Shrink struct task_struct::oomkilladj
oomkilladj is int, but values which can be assigned to it are -17, [-16,
15], thus fitting into s8.
While patch itself doesn't help in making task_struct smaller, because of
natural alignment of ->link_count, it will make picture clearer wrt futher
task_struct reduction patches. My plan is to move ->fpu_counter and
->oomkilladj after ->ioprio filling hole on i386 and x86_64. But that's
for later, because bloated distro configs need looking at as well.
Olaf Hering [Wed, 17 Oct 2007 06:27:13 +0000 (23:27 -0700)]
remove strict ansi check from __u64 in asm/types.h
Remove the __STRICT_ANSI__ check from the __u64/__s64 declaration on
32bit targets.
GCC can be made to warn about usage of long long types with ISO C90
(-ansi), but only with -pedantic. You can write this in a way that even
then it doesn't cause warnings, namely by:
#ifdef __GNUC__
__extension__ typedef __signed__ long long __s64;
__extension__ typedef unsigned long long __u64;
#endif
The __extension__ keyword in front of this switches off any pedantic
warnings for this expression.
Andi Drebes [Wed, 17 Oct 2007 06:27:12 +0000 (23:27 -0700)]
cramfs: error message about endianess
The README file in the cramfs subdirectory says: "All data is currently in
host-endian format; neither mkcramfs nor the kernel ever do swabbing."
If somebody tries to mount a cramfs with the wrong endianess, cramfs only
complains about a wrong magic but doesn't inform the user that only the
endianess isn't right.
The following patch adds an error message to the cramfs sources. If a user
tries to mount a cramfs with the wrong endianess using the patched sources,
cramfs will display the message "cramfs: wrong endianess".
Miklos Szeredi [Wed, 17 Oct 2007 06:27:08 +0000 (23:27 -0700)]
fix execute checking in permission()
permission() checks that MAY_EXEC is only allowed on regular files if at least
one execute bit is set in the file mode.
generic_permission() already ensures this, so the extra check in permission()
is superfluous.
If the filesystem defines it's own ->permission() the check may still be
needed. In this case move it after ->permission(). This is needed because
filesystems such as FUSE may need to refresh the inode attributes before
checking permissions.
This check should be moved inside ->permission(), but that's another story.
Subbaiah Venkata [Wed, 17 Oct 2007 06:27:06 +0000 (23:27 -0700)]
lib/sort.c optimization
Hello, I fixed and tested a small bug in lib/sort.c file, heap sort
function.
The fix avoids unnecessary swap of contents when i is 0 (saves few loads
and stores), which happens every time sort function is called. I felt the
fix is worth bringing it to your attention given the importance and
frequent use of the sort function.
Alan Cox [Wed, 17 Oct 2007 06:27:05 +0000 (23:27 -0700)]
jsm: Remove further unneeded crud
Remove some remaining vestiges of the old hacks jsm had to work around the old
tty buffering. With the new tty buffering it simply doesn't matter any more.
Samuel Thibault [Wed, 17 Oct 2007 06:27:04 +0000 (23:27 -0700)]
unicode diacritics support
There have been issues with non-latin1 diacritics and unicode.
http://bugzilla.kernel.org/show_bug.cgi?id=7746
Git 759448f459234bfcf34b82471f0dba77a9aca498 `Kernel utf-8 handling'
partly resolved it by adding conversion between diacritics and
unicode. The patch below goes further by just turning diacritics into
unicode, hence providing better future support. The kbd support can be
fetched from
http://bugzilla.kernel.org/attachment.cgi?id=12313
This was tested in all of latin1, latin9, latin2 and unicode with french
and czech dead keys.
Turn the kernel accent_table into unicode, and extend ioctls KDGKBDIACR
and KDSKBDIACR into their equivalents KDGKBDIACRUC and KDSKBDIACR.
New function int conv_uni_to_8bit(u32 uni) for converting unicode into 8bit
_input_. No, we don't want to store the translation, as it is potentially
sparse and large.
Ed L. Cashin [Wed, 17 Oct 2007 06:27:03 +0000 (23:27 -0700)]
aoe: remove unecessary wrapper function
We can just use skb_mac_header now, and we don't need a wrapper function to
perform the cast. Instead of requiring the reader to check aoe.h to look
up what an aoe_hdr function does, I'd rather do without it.
When a new block bitmap is read from disk in read_block_bitmap() there are
a few bits that should ALWAYS be set. In particular, the blocks given by
ext4_blk_bitmap, ext4_inode_bitmap and ext4_inode_table. Validate the
block bitmap against these blocks.
projects / linux.git / log