pandora-kernel.git
17 years agoMerge branch 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block
Linus Torvalds [Fri, 11 May 2007 16:58:49 +0000 (09:58 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block:
  Fix compile/link of init/do_mounts.c with !CONFIG_BLOCK
  When stacked block devices are in-use (e.g. md or dm), the recursive calls

17 years agoMerge branch 'audit.b38' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit...
Linus Torvalds [Fri, 11 May 2007 16:57:16 +0000 (09:57 -0700)]
Merge branch 'audit.b38' of git://git./linux/kernel/git/viro/audit-current

* 'audit.b38' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current:
  [PATCH] Abnormal End of Processes
  [PATCH] match audit name data
  [PATCH] complete message queue auditing
  [PATCH] audit inode for all xattr syscalls
  [PATCH] initialize name osid
  [PATCH] audit signal recipients
  [PATCH] add SIGNAL syscall class (v3)
  [PATCH] auditing ptrace

17 years agoMerge branch 'upstream-fixes' of master.kernel.org:/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Fri, 11 May 2007 16:56:05 +0000 (09:56 -0700)]
Merge branch 'upstream-fixes' of /linux/kernel/git/jikos/hid

* 'upstream-fixes' of master.kernel.org:/pub/scm/linux/kernel/git/jikos/hid:
  USB HID: hiddev - fix race between hiddev_send_event() and hiddev_release()
  HID: add hooks for getkeycode() and setkeycode() methods
  HID: switch to using input_dev->dev.parent
  USB HID: Logitech wheel 0x046d/0xc294 needs HID_QUIRK_NOGET quirk
  USB HID: usb_buffer_free() cleanup
  USB HID: report descriptor of Cypress USB barcode readers needs fixup
  Bluetooth HID: HIDP - don't initialize force feedback
  USB HID: update CONFIG_USB_HIDINPUT_POWERBOOK description
  HID: add input mappings for non-working keys on Logitech S510 remote

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Fri, 11 May 2007 16:10:19 +0000 (09:10 -0700)]
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6: (31 commits)
  [NETFILTER]: xt_conntrack: add compat support
  [NETFILTER]: iptable_raw: ignore short packets sent by SOCK_RAW sockets
  [NETFILTER]: iptable_{filter,mangle}: more descriptive "happy cracking" message
  [NETFILTER]: nf_nat: Clears helper private area when NATing
  [NETFILTER]: ctnetlink: clear helper area and handle unchanged helper
  [NETFILTER]: nf_conntrack: Removes unused destroy operation of l3proto
  [NETFILTER]: nf_conntrack: Removes duplicated declarations
  [NETFILTER]: nf_nat: remove unused argument of function allocating binding
  [NETFILTER]: Clean up table initialization
  [NET_SCHED]: Avoid requeue warning on dev_deactivate
  [NET_SCHED]: Reread dev->qdisc for NETDEV_TX_OK
  [NET_SCHED]: Rationalise return value of qdisc_restart
  [NET]: Fix dev->qdisc race for NETDEV_TX_LOCKED case
  [UDP]: Fix AF-specific references in AF-agnostic code.
  [IrDA]: KingSun/DonShine USB IrDA dongle support.
  [IPV6] ROUTE: Assign rt6i_idev for ip6_{prohibit,blk_hole}_entry.
  [IPV6]: Do no rely on skb->dst before it is assigned.
  [IPV6]: Send ICMPv6 error on scope violations.
  [SCTP]: Do not include ABORT chunk header in the notification.
  [SCTP]: Correctly copy addresses in sctp_copy_laddrs
  ...

17 years agoInput: evdev - fix overflow in compat_ioctl
Kenichi Nagai [Fri, 11 May 2007 05:12:15 +0000 (01:12 -0400)]
Input: evdev - fix overflow in compat_ioctl

When exporting input device bitmaps via compat_ioctl on BIG_ENDIAN
platforms evdev calculates data size incorrectly. This causes buffer
overflow if user specifies buffer smaller than maxlen.

Signed-off-by: Kenichi Nagai <kenichi3.nagai@toshiba.co.jp>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomd: improve the is_mddev_idle test
NeilBrown [Fri, 11 May 2007 05:23:31 +0000 (22:23 -0700)]
md: improve the is_mddev_idle test

During a 'resync' or similar activity, md checks if the devices in the
array are otherwise active and winds back resync activity when they are.
This test in done in is_mddev_idle, and it is somewhat fragile - it
sometimes thinks there is non-sync io when there isn't.

The test compares the total sectors of io (disk_stat_read) with the sectors
of resync io (disk->sync_io).  This has problems because total sectors gets
updated when a request completes, while resync io gets updated when the
request is submitted.  The time difference can cause large differenced
between the two which do not actually imply non-resync activity.  The test
currently allows for some fuzz (+/- 4096) but there are some cases when it
is not enough.

The test currently looks for any (non-fuzz) difference, either positive or
negative.  This clearly is not needed.  Any non-sync activity will cause
the total sectors to grow faster than the sync_io count (never slower) so
we only need to look for a positive differences.

If we do this then the amount of in-flight sync io will never cause the
appearance of non-sync IO.  Once enough non-sync IO to worry about starts
happening, resync will be slowed down and the measurements will thus be
more precise (as there is less in-flight) and control of resync will still
be suitably responsive.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoVIDEO: remove archaic if[] construct from Kconfig file
Robert P. J. Day [Fri, 11 May 2007 05:23:30 +0000 (22:23 -0700)]
VIDEO: remove archaic if[] construct from Kconfig file

Remove the obsolete "if [ ]" construct from the video console Kconfig
file.

Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Acked-by: James Simmons <jsimmons@infradead.org>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agopm2fb: fb_sync added
Antonino A. Daplas [Fri, 11 May 2007 05:23:29 +0000 (22:23 -0700)]
pm2fb: fb_sync added

Convert internal wait_pm2() function to fb API fb_sync() method.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Antonino Daplas <adaplas@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agonvidiafb: Enable debugging messages a Kconfig option
Jean Delvare [Fri, 11 May 2007 05:23:29 +0000 (22:23 -0700)]
nvidiafb: Enable debugging messages a Kconfig option

Let the user enable debugging messages in nvidiafb.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Antonino Daplas <adaplas@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agorivafb: Fix I2C getscl callback function
Jean Delvare [Fri, 11 May 2007 05:23:28 +0000 (22:23 -0700)]
rivafb: Fix I2C getscl callback function

Fix rivafb's I2C getscl callback function, as was done in nvidiafb recently.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Antonino Daplas <adaplas@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoatmel_lcdfb: AT91/AT32 LCD Controller framebuffer driver
Nicolas Ferre [Fri, 11 May 2007 05:23:26 +0000 (22:23 -0700)]
atmel_lcdfb: AT91/AT32 LCD Controller framebuffer driver

Adds a framebuffer driver to ATMEL AT91SAM9x and AT32 aka AVR32 platforms.
Those chips share quite the same IP and this code is suitable for both
architectures.

Signed-off-by: Nicolas Ferre <nicolas.ferre@rfo.atmel.com>
Signed-off-by: Antonino Daplas <adaplas@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agopm3fb: Preliminary 2.4 to 2.6 port
Krzysztof Helt [Fri, 11 May 2007 05:23:25 +0000 (22:23 -0700)]
pm3fb: Preliminary 2.4 to 2.6 port

This is a basic port from 2.4 kernel to 2.6.  Acceleration is lost and big
endian support probably too.  The driver works in 8, 16 and 32 bit mode.

[adaplas]
- change VESA_* to FB_BLANK_* constants
- removed unused function clear_memory
- fix uninitialized variable compiler warning
- some whitespace cleaning

[akpm@linux-foundation.org: Nuke pestiferous CVS string]
Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Antonino Daplas <adaplas@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofbdev: geforce 7300 cleanup
Michal Piotrowski [Fri, 11 May 2007 05:23:24 +0000 (22:23 -0700)]
fbdev: geforce 7300 cleanup

ups... coding style.

Signed-off-by: Michal Piotrowski <michal.k.k.piotrowski@gmail.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoepoll cleanups: epoll remove static pre-declarations and akpm-ize the code
Davide Libenzi [Fri, 11 May 2007 05:23:23 +0000 (22:23 -0700)]
epoll cleanups: epoll remove static pre-declarations and akpm-ize the code

Re-arrange epoll code to avoid static functions pre-declarations, and apply
akpm-filter on it.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoepoll cleanups: epoll no module
Davide Libenzi [Fri, 11 May 2007 05:23:22 +0000 (22:23 -0700)]
epoll cleanups: epoll no module

Epoll is either compiled it, or not (if EMBEDDED). Remove the module code
and use fs_initcall().

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoepoll: use anonymous inodes
Davide Libenzi [Fri, 11 May 2007 05:23:21 +0000 (22:23 -0700)]
epoll: use anonymous inodes

Cut out lots of code from epoll, by reusing the anonymous inode source
patch (fs/anon_inodes.c).

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: KAIO eventfd support example
Davide Libenzi [Fri, 11 May 2007 05:23:21 +0000 (22:23 -0700)]
signal/timer/event: KAIO eventfd support example

This is an example about how to add eventfd support to the current KAIO code,
in order to enable KAIO to post readiness events to a pollable fd (hence
compatible with POSIX select/poll).  The KAIO code simply signals the eventfd
fd when events are ready, and this triggers a POLLIN in the fd.  This patch
uses a reserved for future use member of the struct iocb to pass an eventfd
file descriptor, that KAIO will use to post events every time a request
completes.  At that point, an aio_getevents() will return the completed result
to a struct io_event.  I made a quick test program to verify the patch, and it
runs fine here:

http://www.xmailserver.org/eventfd-aio-test.c

The test program uses poll(2), but it'd, of course, work with select and epoll
too.

This can allow to schedule both block I/O and other poll-able devices
requests, and wait for results using select/poll/epoll.  In a typical
scenario, an application would submit KAIO request using aio_submit(), and
will also use epoll_ctl() on the whole other class of devices (that with the
addition of signals, timers and user events, now it's pretty much complete),
and then would:

epoll_wait(...);
for_each_event {
if (curr_event_is_kaiofd) {
aio_getevents();
dispatch_aio_events();
} else {
dispatch_epoll_event();
}
}

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: eventfd wire up x86 arches
Davide Libenzi [Fri, 11 May 2007 05:23:20 +0000 (22:23 -0700)]
signal/timer/event: eventfd wire up x86 arches

This patch wires the eventfd system call to the x86 architectures.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: eventfd core
Davide Libenzi [Fri, 11 May 2007 05:23:19 +0000 (22:23 -0700)]
signal/timer/event: eventfd core

This is a very simple and light file descriptor, that can be used as event
wait/dispatch by userspace (both wait and dispatch) and by the kernel
(dispatch only).  It can be used instead of pipe(2) in all cases where those
would simply be used to signal events.  Their kernel overhead is much lower
than pipes, and they do not consume two fds.  When used in the kernel, it can
offer an fd-bridge to enable, for example, functionalities like KAIO or
syslets/threadlets to signal to an fd the completion of certain operations.
But more in general, an eventfd can be used by the kernel to signal readiness,
in a POSIX poll/select way, of interfaces that would otherwise be incompatible
with it.  The API is:

int eventfd(unsigned int count);

The eventfd API accepts an initial "count" parameter, and returns an eventfd
fd.  It supports poll(2) (POLLIN, POLLOUT, POLLERR), read(2) and write(2).

The POLLIN flag is raised when the internal counter is greater than zero.

The POLLOUT flag is raised when at least a value of "1" can be written to the
internal counter.

The POLLERR flag is raised when an overflow in the counter value is detected.

The write(2) operation can never overflow the counter, since it blocks (unless
O_NONBLOCK is set, in which case -EAGAIN is returned).

But the eventfd_signal() function can do it, since it's supposed to not sleep
during its operation.

The read(2) function reads the __u64 counter value, and reset the internal
value to zero.  If the value read is equal to (__u64) -1, an overflow happened
on the internal counter (due to 2^64 eventfd_signal() posts that has never
been retired - unlickely, but possible).

The write(2) call writes an __u64 count value, and adds it to the current
counter.  The eventfd fd supports O_NONBLOCK also.

On the kernel side, we have:

struct file *eventfd_fget(int fd);
int eventfd_signal(struct file *file, unsigned int n);

The eventfd_fget() should be called to get a struct file* from an eventfd fd
(this is an fget() + check of f_op being an eventfd fops pointer).

The kernel can then call eventfd_signal() every time it wants to post an event
to userspace.  The eventfd_signal() function can be called from any context.
An eventfd() simple test and bench is available here:

http://www.xmailserver.org/eventfd-bench.c

This is the eventfd-based version of pipetest-4 (pipe(2) based):

http://www.xmailserver.org/pipetest-4.c

Not that performance matters much in the eventfd case, but eventfd-bench
shows almost as double as performance than pipetest-4.

[akpm@linux-foundation.org: fix i386 build]
[akpm@linux-foundation.org: add sys_eventfd to sys_ni.c]
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: timerfd compat code
Davide Libenzi [Fri, 11 May 2007 05:23:18 +0000 (22:23 -0700)]
signal/timer/event: timerfd compat code

This patch implements the necessary compat code for the timerfd system call.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: timerfd wire up x86 arches
Davide Libenzi [Fri, 11 May 2007 05:23:17 +0000 (22:23 -0700)]
signal/timer/event: timerfd wire up x86 arches

This patch wires the timerfd system call to the x86 architectures.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Andi Kleen <ak@suse.de>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: timerfd core
Davide Libenzi [Fri, 11 May 2007 05:23:16 +0000 (22:23 -0700)]
signal/timer/event: timerfd core

This patch introduces a new system call for timers events delivered though
file descriptors.  This allows timer event to be used with standard POSIX
poll(2), select(2) and read(2).  As a consequence of supporting the Linux
f_op->poll subsystem, they can be used with epoll(2) too.

The system call is defined as:

int timerfd(int ufd, int clockid, int flags, const struct itimerspec *utmr);

The "ufd" parameter allows for re-use (re-programming) of an existing timerfd
w/out going through the close/open cycle (same as signalfd).  If "ufd" is -1,
s new file descriptor will be created, otherwise the existing "ufd" will be
re-programmed.

The "clockid" parameter is either CLOCK_MONOTONIC or CLOCK_REALTIME.  The time
specified in the "utmr->it_value" parameter is the expiry time for the timer.

If the TFD_TIMER_ABSTIME flag is set in "flags", this is an absolute time,
otherwise it's a relative time.

If the time specified in the "utmr->it_interval" is not zero (.tv_sec == 0,
tv_nsec == 0), this is the period at which the following ticks should be
generated.

The "utmr->it_interval" should be set to zero if only one tick is requested.
Setting the "utmr->it_value" to zero will disable the timer, or will create a
timerfd without the timer enabled.

The function returns the new (or same, in case "ufd" is a valid timerfd
descriptor) file, or -1 in case of error.

As stated before, the timerfd file descriptor supports poll(2), select(2) and
epoll(2).  When a timer event happened on the timerfd, a POLLIN mask will be
returned.

The read(2) call can be used, and it will return a u32 variable holding the
number of "ticks" that happened on the interface since the last call to
read(2).  The read(2) call supportes the O_NONBLOCK flag too, and EAGAIN will
be returned if no ticks happened.

A quick test program, shows timerfd working correctly on my amd64 box:

http://www.xmailserver.org/timerfd-test.c

[akpm@linux-foundation.org: add sys_timerfd to sys_ni.c]
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: signalfd compat code
Davide Libenzi [Fri, 11 May 2007 05:23:15 +0000 (22:23 -0700)]
signal/timer/event: signalfd compat code

This patch implements the necessary compat code for the signalfd system call.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: signalfd wire up x86 arches
Davide Libenzi [Fri, 11 May 2007 05:23:14 +0000 (22:23 -0700)]
signal/timer/event: signalfd wire up x86 arches

This patch wires the signalfd system call to the x86 architectures.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event: signalfd core
Davide Libenzi [Fri, 11 May 2007 05:23:13 +0000 (22:23 -0700)]
signal/timer/event: signalfd core

This patch series implements the new signalfd() system call.

I took part of the original Linus code (and you know how badly it can be
broken :), and I added even more breakage ;) Signals are fetched from the same
signal queue used by the process, so signalfd will compete with standard
kernel delivery in dequeue_signal().  If you want to reliably fetch signals on
the signalfd file, you need to block them with sigprocmask(SIG_BLOCK).  This
seems to be working fine on my Dual Opteron machine.  I made a quick test
program for it:

http://www.xmailserver.org/signafd-test.c

The signalfd() system call implements signal delivery into a file descriptor
receiver.  The signalfd file descriptor if created with the following API:

int signalfd(int ufd, const sigset_t *mask, size_t masksize);

The "ufd" parameter allows to change an existing signalfd sigmask, w/out going
to close/create cycle (Linus idea).  Use "ufd" == -1 if you want a brand new
signalfd file.

The "mask" allows to specify the signal mask of signals that we are interested
in.  The "masksize" parameter is the size of "mask".

The signalfd fd supports the poll(2) and read(2) system calls.  The poll(2)
will return POLLIN when signals are available to be dequeued.  As a direct
consequence of supporting the Linux poll subsystem, the signalfd fd can use
used together with epoll(2) too.

The read(2) system call will return a "struct signalfd_siginfo" structure in
the userspace supplied buffer.  The return value is the number of bytes copied
in the supplied buffer, or -1 in case of error.  The read(2) call can also
return 0, in case the sighand structure to which the signalfd was attached,
has been orphaned.  The O_NONBLOCK flag is also supported, and read(2) will
return -EAGAIN in case no signal is available.

If the size of the buffer passed to read(2) is lower than sizeof(struct
signalfd_siginfo), -EINVAL is returned.  A read from the signalfd can also
return -ERESTARTSYS in case a signal hits the process.  The format of the
struct signalfd_siginfo is, and the valid fields depends of the (->code &
__SI_MASK) value, in the same way a struct siginfo would:

struct signalfd_siginfo {
__u32 signo; /* si_signo */
__s32 err; /* si_errno */
__s32 code; /* si_code */
__u32 pid; /* si_pid */
__u32 uid; /* si_uid */
__s32 fd; /* si_fd */
__u32 tid; /* si_fd */
__u32 band; /* si_band */
__u32 overrun; /* si_overrun */
__u32 trapno; /* si_trapno */
__s32 status; /* si_status */
__s32 svint; /* si_int */
__u64 svptr; /* si_ptr */
__u64 utime; /* si_utime */
__u64 stime; /* si_stime */
__u64 addr; /* si_addr */
};

[akpm@linux-foundation.org: fix signalfd_copyinfo() on i386]
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosignal/timer/event fds: anonymous inode source
Davide Libenzi [Fri, 11 May 2007 05:23:11 +0000 (22:23 -0700)]
signal/timer/event fds: anonymous inode source

This patch add an anonymous inode source, to be used for files that need
and inode only in order to create a file*. We do not care of having an
inode for each file, and we do not even care of having different names in
the associated dentries (dentry names will be same for classes of file*).
This allow code reuse, and will be used by epoll, signalfd and timerfd
(and whatever else there'll be).

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoDon't init pgrp and __session in INIT_SIGNALS
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:10 +0000 (22:23 -0700)]
Don't init pgrp and __session in INIT_SIGNALS

Remove initialization of pgrp and __session in INIT_SIGNALS, as these are
later set by the call to __set_special_pids() in init/main.c by the patch:

explicitly-set-pgid-and-sid-of-init-process.patch

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoReplace pid_t in autofs with struct pid reference
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:08 +0000 (22:23 -0700)]
Replace pid_t in autofs with struct pid reference

Make autofs container-friendly by caching struct pid reference rather than
pid_t and using pid_nr() to retreive a task's pid_t.

ChangeLog:
- Fix Eric Biederman's comments - Use find_get_pid() to hold a
  reference to oz_pgrp and release while unmounting; separate out
  changes to autofs and autofs4.
- Fix Cedric's comments: retain old prototype of parse_options()
  and move necessary change to its caller.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: containers@lists.osdl.org
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoFix some coding-style errors in autofs
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:06 +0000 (22:23 -0700)]
Fix some coding-style errors in autofs

Fix coding style errors (extra spaces, long lines) in autofs and autofs4 files
being modified for container/pidspace issues.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: <containers@lists.osdl.org>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoKill unused sesssion and group values in rocket driver
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:05 +0000 (22:23 -0700)]
Kill unused sesssion and group values in rocket driver

The process_session() and process_group() values are not really used by the
driver.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: <containers@lists.osdl.org>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoUse task_pgrp() task_session() in copy_process()
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:04 +0000 (22:23 -0700)]
Use task_pgrp() task_session() in copy_process()

Use task_pgrp() and task_session() in copy_process(), and avoid find_pid()
call when attaching the task to its process group and session.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: <containers@lists.osdl.org>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoUse struct pid parameter in copy_process()
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:03 +0000 (22:23 -0700)]
Use struct pid parameter in copy_process()

Modify copy_process() to take a struct pid * parameter instead of a pid_t.
This simplifies the code a bit and also avoids having to call find_pid() to
convert the pid_t to a struct pid.

Changelog:
- Fixed Badari Pulavarty's comments and passed in &init_struct_pid
  from fork_idle().
- Fixed Eric Biederman's comments and simplified this patch and
  used a new patch to remove the likely(pid) check.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: <containers@lists.osdl.org>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoExplicitly set pgid and sid of init process
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:01 +0000 (22:23 -0700)]
Explicitly set pgid and sid of init process

Explicitly set pgid and sid of init process to 1.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: <containers@lists.osdl.org>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agostatically initialize struct pid for swapper
Sukadev Bhattiprolu [Fri, 11 May 2007 05:23:00 +0000 (22:23 -0700)]
statically initialize struct pid for swapper

Statically initialize a struct pid for the swapper process (pid_t == 0) and
attach it to init_task.  This is needed so task_pid(), task_pgrp() and
task_session() interfaces work on the swapper process also.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: <containers@lists.osdl.org>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoattach_pid() with struct pid parameter
Sukadev Bhattiprolu [Fri, 11 May 2007 05:22:58 +0000 (22:22 -0700)]
attach_pid() with struct pid parameter

attach_pid() currently takes a pid_t and then uses find_pid() to find the
corresponding struct pid.  Sometimes we already have the struct pid.  We can
then skip find_pid() if attach_pid() were to take a struct pid parameter.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: <containers@lists.osdl.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agortc-rs5c313.c: add error handling to avoid hardware hangup
kogiidena [Fri, 11 May 2007 05:22:57 +0000 (22:22 -0700)]
rtc-rs5c313.c: add error handling to avoid hardware hangup

Add error processing.  Hanging up by an infinite loop is evaded.

Signed-off-by: kogiidena <kogiidena@eggplant.ddo.jp>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agortc-rs5c313.c: rtc_time value are fixed
kogiidena [Fri, 11 May 2007 05:22:56 +0000 (22:22 -0700)]
rtc-rs5c313.c: rtc_time value are fixed

Correct an initial value of suruct rtc_ time.

Signed-off-by: kogiidena <kogiidena@eggplant.ddo.jp>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agortc-rs5c313.c: error and warning are fixed
kogiidena [Fri, 11 May 2007 05:22:54 +0000 (22:22 -0700)]
rtc-rs5c313.c: error and warning are fixed

Correct a compile error and warning.

Signed-off-by: kogiidena <kogiidena@eggplant.ddo.jp>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouse defines in sys_getpriority/sys_setpriority
Daniel Walker [Fri, 11 May 2007 05:22:53 +0000 (22:22 -0700)]
use defines in sys_getpriority/sys_setpriority

Switch to the defines for these two checks, instead of hard coding the
values.

[akpm@linux-foundation.org: add missing include]
Signed-off-by: Daniel Walker <dwalker@mvista.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMPC52xx PSC SPI master driver
Dragos Carp [Fri, 11 May 2007 05:22:52 +0000 (22:22 -0700)]
MPC52xx PSC SPI master driver

SPI master driver for MPC52xx using its Programmable Serial Controller.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Dragos Carp <dragos.carp@toptica.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoconsolidate generic_writepages and mpage_writepages
Miklos Szeredi [Fri, 11 May 2007 05:22:51 +0000 (22:22 -0700)]
consolidate generic_writepages and mpage_writepages

Clean up massive code duplication between mpage_writepages() and
generic_writepages().

The new generic function, write_cache_pages() takes a function pointer
argument, which will be called for each page to be written.

Maybe cifs_writepages() too can use this infrastructure, but I'm not
touching that with a ten-foot pole.

The upcoming page writeback support in fuse will also want this.

Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acked-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agotty: add compat_ioctl
Paul Fulghum [Fri, 11 May 2007 05:22:50 +0000 (22:22 -0700)]
tty: add compat_ioctl

Add compat_ioctl method for tty code to allow processing of 32 bit ioctl
calls on 64 bit systems by tty core, tty drivers, and line disciplines.

Based on patch by Arnd Bergmann:
http://www.uwsg.iu.edu/hypermail/linux/kernel/0511.0/1732.html

[akpm@linux-foundation.org: make things static]
Signed-off-by: Paul Fulghum <paulkf@microgate.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomodule_author: don't advise putting in an email address
Rene Herman [Fri, 11 May 2007 05:22:50 +0000 (22:22 -0700)]
module_author: don't advise putting in an email address

module_author: don't advise putting in an email address

It's information that's easily outdated and easily mistaken for a driver
contact which is a problem especially for modules with multiple current and
non-current authors as well as for modules with a maintainer who may not
even be a module author.

Signed-off-by: Rene Herman <rene.herman@gmail.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoSubmitChecklist: add -W help
Andrew Morton [Fri, 11 May 2007 05:22:49 +0000 (22:22 -0700)]
SubmitChecklist: add -W help

Help people to work out how to use `gcc -W'.

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoOverrun in drivers/char/rio/riocmd.c
Eric Sesterhenn / Snakebyte [Fri, 11 May 2007 05:22:48 +0000 (22:22 -0700)]
Overrun in drivers/char/rio/riocmd.c

This got somehow lost in the noise.  This fixes coverity bug id #1025, if
Rup is greater or equal to MAX_RUP, we run past the Mapping Array.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agostop_machine() now uses hard_irq_disable
Benjamin Herrenschmidt [Fri, 11 May 2007 05:22:47 +0000 (22:22 -0700)]
stop_machine() now uses hard_irq_disable

Add a call to hard_irq_disable() to stop_machine so that we make sure IRQs are
really disabled and not only lazy-disabled on archs like powerpc as some users
of stop_machine() may rely on that.

[akpm@linux-foundation.org: build fix]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoAdd hard_irq_disable()
Benjamin Herrenschmidt [Fri, 11 May 2007 05:22:46 +0000 (22:22 -0700)]
Add hard_irq_disable()

Some architectures, like powerpc, implement lazy disabling of interrupts.
That means that on those, local_irq_disable() doesn't actually disable
interrupts on the CPU, but only sets some per CPU flag which cause them to be
disabled only if an interrupt actually occurs.

However, in some cases, such as stop_machine, we really want interrupts to be
fully disabled.  For example, I have code using stop machine to do ECC error
injection, used to verify operations of the ECC hardware, that sort of thing.
It really needs to make sure that nothing is actually writing to memory while
the injection happens.  Similar examples can be found in other low level bits
and pieces.

This patch implements a generic hard_irq_disable() function which is meant to
be called -after- local_irq_disable() and ensures that interrupts are fully
disabled on that CPU.  The default implementation is a nop, though powerpc
does already provide an appropriate one.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agopowerpc: fixup hard_irq_disable semantics
Benjamin Herrenschmidt [Fri, 11 May 2007 05:22:45 +0000 (22:22 -0700)]
powerpc: fixup hard_irq_disable semantics

This patch renames the raw hard_irq_{enable,disable} into
__hard_irq_{enable,disable} and introduces a higher level hard_irq_disable()
function that can be used by any code to enforce that IRQs are fully disabled,
not only lazy disabled.

The difference with the __ versions is that it will update some per-processor
fields so that the kernel keeps track and properly re-enables them in the next
local_irq_disable();

This prepares powerpc for my next patch that introduces hard_irq_disable()
generically.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosynclink_gt: add compat_ioctl
Paul Fulghum [Fri, 11 May 2007 05:22:43 +0000 (22:22 -0700)]
synclink_gt: add compat_ioctl

Add support for 32 bit ioctl on 64 bit systems for synclink_gt

Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul Fulghum <paulkf@microgate.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosmall cleanup in gpt partition handling
Olaf Hering [Fri, 11 May 2007 05:22:42 +0000 (22:22 -0700)]
small cleanup in gpt partition handling

Remove unused argument in is_pmbr_valid()
Remove unneeded initialization of local variable legacy_mbr

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoConsolidate asm/poll.h
Stephen Rothwell [Fri, 11 May 2007 05:22:40 +0000 (22:22 -0700)]
Consolidate asm/poll.h

These files are almost all the same.

This patch could be made even simpler if we don't mind POLLREMOVE turning
up in a few architectures that didn't have it previously (which should be
OK as POLLREMOVE is not used anywhere in the current tree).

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolib/hexdump
Randy Dunlap [Fri, 11 May 2007 05:22:39 +0000 (22:22 -0700)]
lib/hexdump

Based on ace_dump_mem() from Grant Likely for the Xilinx SystemACE
CompactFlash interface.

Add print_hex_dump() & hex_dumper() to lib/hexdump.c and linux/kernel.h.

This patch adds the functions print_hex_dump() & hex_dumper().
print_hex_dump() can be used to perform a hex + ASCII dump of data to
syslog, in an easily viewable format, thus providing a common text hex dump
format.

hex_dumper() provides a dump-to-memory function.  It converts one "line" of
output (16 bytes of input) at a time.

Example usages:
print_hex_dump(KERN_DEBUG, DUMP_PREFIX_ADDRESS, frame->data, frame->len);
hex_dumper(frame->data, frame->len, linebuf, sizeof(linebuf));

Example output using %DUMP_PREFIX_OFFSET:
0009ab4240414243 44454647 48494a4b 4c4d4e4f-@ABCDEFG HIJKLMNO
Example output using %DUMP_PREFIX_ADDRESS:
ffffffff88089af070717273 74757677 78797a7b 7c7d7e7f-pqrstuvw xyz{|}~.

[akpm@linux-foundation.org: cleanups, add export]
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agogetrusage(): fill ru_inblock and ru_oublock fields if possible
Eric Dumazet [Fri, 11 May 2007 05:22:37 +0000 (22:22 -0700)]
getrusage(): fill ru_inblock and ru_oublock fields if possible

If CONFIG_TASK_IO_ACCOUNTING is defined, we update io accounting counters for
each task.

This patch permits reporting of values using the well known getrusage()
syscall, filling ru_inblock and ru_oublock instead of null values.

As TASK_IO_ACCOUNTING currently counts bytes counts, we approximate blocks
count doing : nr_blocks = nr_bytes / 512

Example of use :
----------------------
After patch is applied, /usr/bin/time command can now give a good
approximation of IO that the process had to do.

$ /usr/bin/time grep tototo /usr/include/*
Command exited with non-zero status 1
0.00user 0.02system 0:02.11elapsed 1%CPU (0avgtext+0avgdata 0maxresident)k
24288inputs+0outputs (0major+259minor)pagefaults 0swaps

$ /usr/bin/time dd if=/dev/zero of=/tmp/testfile count=1000
1000+0 enregistrements lus
1000+0 enregistrements écrits
512000 octets (512 kB) copiés, 0,00326601 seconde, 157 MB/s
0.00user 0.00system 0:00.00elapsed 80%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+3000outputs (0major+299minor)pagefaults 0swaps

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouml: shrink kernel stacks
Jeff Dike [Fri, 11 May 2007 05:22:35 +0000 (22:22 -0700)]
uml: shrink kernel stacks

Make kernel stacks be 1 page on i386 and 2 pages on x86_64.  These match the
host values.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouml: iRQ stacks
Jeff Dike [Fri, 11 May 2007 05:22:34 +0000 (22:22 -0700)]
uml: iRQ stacks

Add a separate IRQ stack.  This differs from i386 in having the entire
interrupt run on a separate stack rather than starting on the normal kernel
stack and switching over once some preparation has been done.  The underlying
mechanism, is of course, sigaltstack.

Another difference is that interrupts that happen in userspace are handled on
the normal kernel stack.  These cause a wait wakeup instead of a signal
delivery so there is no point in trying to switch stacks for these.  There's
no other stuff on the stack, so there is no extra stack consumption.

This quirk makes it possible to have the entire interrupt run on a separate
stack - process preemption (and calls to schedule()) happens on a normal
kernel stack.  If we enable CONFIG_PREEMPT, this will need to be rethought.

The IRQ stack for CPU 0 is declared in the same way as the initial kernel
stack.  IRQ stacks for other CPUs will be allocated dynamically.

An extra field was added to the thread_info structure.  When the active
thread_info is copied to the IRQ stack, the real_thread field points back to
the original stack.  This makes it easy to tell where to copy the thread_info
struct back to when the interrupt is finished.  It also serves as a marker of
a nested interrupt.  It is NULL for the first interrupt on the stack, and
non-NULL for any nested interrupts.

Care is taken to behave correctly if a second interrupt comes in when the
thread_info structure is being set up or taken down.  I could just disable
interrupts here, but I don't feel like giving up any of the performance gained
by not flipping signals on and off.

If an interrupt comes in during these critical periods, the handler can't run
because it has no idea what shape the stack is in.  So, it sets a bit for its
signal in a global mask and returns.  The outer handler will deal with this
signal itself.

Atomicity is had with xchg.  A nested interrupt that needs to bail out will
xchg its signal mask into pending_mask and repeat in case yet another
interrupt hit at the same time, until the mask stabilizes.

The outermost interrupt will set up the thread_info and xchg a zero into
pending_mask when it is done.  At this point, nested interrupts will look at
->real_thread and see that no setup needs to be done.  They can just continue
normally.

Similar care needs to be taken when exiting the outer handler.  If another
interrupt comes in while it is copying the thread_info, it will drop a bit
into pending_mask.  The outer handler will check this and if it is non-zero,
will loop, set up the stack again, and handle the interrupt.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouml: tidy IRQ code
Jeff Dike [Fri, 11 May 2007 05:22:32 +0000 (22:22 -0700)]
uml: tidy IRQ code

Some tidying of the irq code before introducing irq stacks.  Mostly
style fixes, but the timer handler calls the timer code directly
rather than going through the generic sig_handler_common_skas.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouml: use UM_THREAD_SIZE in userspace code
Jeff Dike [Fri, 11 May 2007 05:22:31 +0000 (22:22 -0700)]
uml: use UM_THREAD_SIZE in userspace code

Now that we have UM_THREAD_SIZE, we can replace the calculations in
user-space code (an earlier patch took care of the kernel side of the
house).

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouml: remove task_protections
Jeff Dike [Fri, 11 May 2007 05:22:30 +0000 (22:22 -0700)]
uml: remove task_protections

Replaced task_protections with stack_protections since they do the same
thing, and task_protections was misnamed anyway.

This needs THREAD_SIZE, so that's imported via common-offsets.h

Also tidied up the code in the vicinity.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoLet SYSV68_PARTITION default to yes on VME only
Geert Uytterhoeven [Fri, 11 May 2007 05:22:28 +0000 (22:22 -0700)]
Let SYSV68_PARTITION default to yes on VME only

Don't enable SYSV68 partition table support on all m68k boxes by default,
only on Motorola VME boards.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Philippe De Muyter <phdm@macqel.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom32r: fix pte_to_pgoff(), pgoff_to_pte() and __swp_type() macros
Hirokazu Takata [Fri, 11 May 2007 05:22:28 +0000 (22:22 -0700)]
m32r: fix pte_to_pgoff(), pgoff_to_pte() and __swp_type() macros

This patch is required to handle file-mapped or swapped-out pages
correctly.

- Fix pte_to_pgoff() and pgoff_to_pte() macros not to include
  _PAGE_PROTNONE bit of PTE.
  Mask value for { ACCESSED, N, (R, W, X), L, G } is not 0xef but 0x7f.
- Fix __swp_type() macro for MAX_SWAPFILES_SHIFT(=5), which is defined
  in include/linux/swap.h.

* M32R TLB format

     [0]    [1:19]           [20:23]       [24:31]
     +-----------------------+----+-------------+
     |          VPN          |0000|    ASID     |
     +-----------------------+----+-------------+
     +-+---------------------+----+-+---+-+-+-+-+
     |0         PPN          |0000|N|AC |L|G|V| |
     +-+---------------------+----+-+---+-+-+-+-+
                                ||   RWX     | |
* software bits in PTE          ||           | +-- _PAGE_FILE | _PAGE_DIRTY
                                ||           +---- _PAGE_PRESENT
                                |+---------------- _PAGE_ACCESSED
                                +----------------- _PAGE_PROTNONE

Signed-off-by: Hitoshi Yamamoto <hitoshiy@linux-m32r.org>
Signed-off-by: Hirokazu Takata <takata@linux-m32r.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom32r: fix tme_handler to check _PAGE_PRESENT bit
Hirokazu Takata [Fri, 11 May 2007 05:22:26 +0000 (22:22 -0700)]
m32r: fix tme_handler to check _PAGE_PRESENT bit

Fix the tlb-miss handler (tme_handler) to check _PAGE_PRESENT bit
in order to handle file-mapped or swapped-out pages correctly.

This patch is required to fix unexpected page errors for m32r.

Signed-off-by: Hitoshi Yamamoto <hitoshiy@linux-m32r.org>
Signed-off-by: Hirokazu Takata <takata@linux-m32r.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom32r: fix switch_to macro to push/pop frame pointer if needed
Hirokazu Takata [Fri, 11 May 2007 05:22:25 +0000 (22:22 -0700)]
m32r: fix switch_to macro to push/pop frame pointer if needed

This patch fixes a rarely-happened but severe scheduling problem of
the recent m32r kernel of 2.6.17-rc3 or later.

In the following previous m32r patch, the switch_to macro was
modified not to do unnecessary push/pop operations for tuning.
> [PATCH] m32r: update switch_to macro for tuning
4127272c38619c56f0c1aa01d01c7bd757db70a1

In this modification, only 'lr' and 'sp' registers are push/pop'ed,
assuming that the m32r kernel is always compiled with
-fomit-frame-pointer option.

However, in 2.6 kernel, kernel/sched.c is irregularly compiled
with -fno-omit-frame-pointer if CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER
is not defined.

 -- kernel/Makefile --
   :
 ifneq ($(CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER),y)
 # According to Alan Modra <alan@linuxcare.com.au>, the -fno-omit-frame-pointer is
 # needed for x86 only.  Why this used to be enabled for all architectures is beyond
 # me.  I suspect most platforms don't need this, but until we know that for sure
 # I turn this off for IA-64 only.  Andreas Schwab says it's also needed on m68k
 # to get a correct value for the wait-channel (WCHAN in ps). --davidm
 CFLAGS_sched.o := $(PROFILING) -fno-omit-frame-pointer
 endif
   :
 ---

Therefore, for the recent m32r kernel, we have to push/pop 'fp'
(frame pointer) if CONFIG_FRAME_POINTER is defined or
CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER is not defined.

Signed-off-by: Hitoshi Yamamoto <hitoshiy@linux-m32r.org>
Signed-off-by: Hirokazu Takata <takata@linux-m32r.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoh8300 syscall update
Yoshinori Sato [Fri, 11 May 2007 05:22:23 +0000 (22:22 -0700)]
h8300 syscall update

h8300 systemcall entry table update.

Signed-off-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofrv: gdb: use __maybe_unused
David Rientjes [Fri, 11 May 2007 05:22:22 +0000 (22:22 -0700)]
frv: gdb: use __maybe_unused

Replace function instances of __attribute__((unused)) with
__maybe_unused to suppress warnings.

Cc: David Howells <dhowells@redhat.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoVM statistics: Make timer deferrable
Christoph Lameter [Fri, 11 May 2007 05:22:21 +0000 (22:22 -0700)]
VM statistics: Make timer deferrable

VM statistics updates do not matter if the kernel is in idle powersaving
mode.  So allow the timer to be deferred.

It would be better though if we could switch the timer between deferrable
and nondeferrable based on differentials present.  The timer would start
out nondeferrable and if we find that there were no updates in the last
statistics interval then we would switch the timer to deferrable.  If the
timer later finds again that there are differentials then go to
nondeferrable again.

And yet another way would be to run the timer shortly before going to idle?

The solution here means that the VM counters may be slightly off during
idle since differentials may be still pending while the timer is deferred.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoAFS: implement statfs
David Howells [Fri, 11 May 2007 05:22:20 +0000 (22:22 -0700)]
AFS: implement statfs

Implement the statfs() op for AFS.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoAFS: fix a couple of problems with unlinking AFS files
David Howells [Fri, 11 May 2007 05:22:20 +0000 (22:22 -0700)]
AFS: fix a couple of problems with unlinking AFS files

Fix a couple of problems with unlinking AFS files.

 (1) The parent directory wasn't being updated properly between unlink() and
     the following lookup().

     It seems that, for some reason, invalidate_remote_inode() wasn't
     discarding the directory contents correctly, so this patch calls
     invalidate_inode_pages2() instead on non-regular files.

 (2) afs_vnode_deleted_remotely() should handle vnodes that don't have a
     source server recorded without oopsing.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoAFS: fix interminable loop in afs_write_back_from_locked_page()
David Howells [Fri, 11 May 2007 05:22:19 +0000 (22:22 -0700)]
AFS: fix interminable loop in afs_write_back_from_locked_page()

Following bug was uncovered by compiling with '-W' flag:

  CC [M]  fs/afs/write.o
fs/afs/write.c: In function â\80\98afs_write_back_from_locked_pageâ\80\99:
fs/afs/write.c:398: warning: comparison of unsigned expression >= 0 is always true

Loop variable 'n' is unsigned, so wraps around happily as far as I can
see. Trival fix attached (compile tested only).

Signed-off-by: Mika Kukkonen <mikukkon@iki.fi>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: new syscall
Andi Kleen [Fri, 11 May 2007 05:22:18 +0000 (22:22 -0700)]
x86_64: new syscall

Add epoll_pwait()

(akpm: stolen from Andi's queue, because I want to send the signalfd patches
which also add syscalls.  Not sure what the __IGNORE_getcpu is for).

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoDocumentation/gpio.txt mentions GENERIC_GPIO
David Brownell [Fri, 11 May 2007 05:22:17 +0000 (22:22 -0700)]
Documentation/gpio.txt mentions GENERIC_GPIO

Documentation/gpio.txt should mention the Kconfig GENERIC_GPIO flag, for
platforms to declare when relevant.  This should help minimize goofs like
omitting it, or not depending on it when needed.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoBug in mm/thrash.c function grab_swap_token()
Mika Kukkonen [Fri, 11 May 2007 05:22:17 +0000 (22:22 -0700)]
Bug in mm/thrash.c function grab_swap_token()

Following bug was uncovered by compiling with '-W' flag:

  CC      mm/thrash.o
mm/thrash.c: In function â\80\98grab_swap_tokenâ\80\99:
mm/thrash.c:52: warning: comparison of unsigned expression < 0 is always false

Variable token_priority is unsigned, so decrementing first and then
checking the result does not work; fixed by reversing the test, patch
attached (compile tested only).

I am not sure if likely() makes much sense in this new situation, but
I'll let somebody else to make a decision on that.

Signed-off-by: Mika Kukkonen <mikukkon@iki.fi>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: display more intuitive error message if kernel is not 2MB aligned
Vivek Goyal [Fri, 11 May 2007 05:22:15 +0000 (22:22 -0700)]
x86_64: display more intuitive error message if kernel is not 2MB aligned

o x86_64 kernel needs to be compiled for 2MB aligned addresses. Currently
  we are using BUILD_BUG_ON() to warn the user if he has not done so. But
  looks like folks are not finding message very intutive and don't open
  the respective c file to find problem source. (Bug 8439)

arch/x86_64/kernel/head64.c: In function 'x86_64_start_kernel':
arch/x86_64/kernel/head64.c:70: error: size of array 'type name' is negative

o Using preprocessor directive #error to print a better message if
  CONFIG_PHYSICAL_START is not aligned to 2MB boundary.

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Cc: Andi Kleen <ak@suse.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoi386: work around miscompilation of alternatives code
Joerg Roedel [Fri, 11 May 2007 05:22:14 +0000 (22:22 -0700)]
i386: work around miscompilation of alternatives code

A recent change makes my Dell 1501 hang on boot.  It's an AMD MK-36.  I use
an x86_64 kernel.  It is 100% reproducible.

I debugged this problem a bit and my compiler[1]interprets the =A constraint
as %rax instead of %edx:%eax on x86_64 which causes the problem.  The appended
patch provides a workaround for this and fixed the hang on my machine.

[1] gcc version 4.1.3 20070429 (prerelease) (Debian 4.1.2-5)

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Benny Halevy <bhalevy@panasas.com>
Cc: Pete Zaitcev <zaitcev@redhat.com>
Cc: "Joerg Roedel" <joerg.roedel@amd.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoFix compile/link of init/do_mounts.c with !CONFIG_BLOCK
Jens Axboe [Fri, 11 May 2007 11:29:54 +0000 (13:29 +0200)]
Fix compile/link of init/do_mounts.c with !CONFIG_BLOCK

We need a stub function for when CONFIG_BLOCK isn't set.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
17 years agoWhen stacked block devices are in-use (e.g. md or dm), the recursive calls
Neil Brown [Tue, 1 May 2007 07:53:42 +0000 (09:53 +0200)]
When stacked block devices are in-use (e.g. md or dm), the recursive calls

to generic_make_request can use up a lot of space, and we would rather they
didn't.

As generic_make_request is a void function, and as it is generally not
expected that it will have any effect immediately, it is safe to delay any
call to generic_make_request until there is sufficient stack space
available.

As ->bi_next is reserved for the driver to use, it can have no valid value
when generic_make_request is called, and as __make_request implicitly
assumes it will be NULL (ELEVATOR_BACK_MERGE fork of switch) we can be
certain that all callers set it to NULL.  We can therefore safely use
bi_next to link pending requests together, providing we clear it before
making the real call.

So, we choose to allow each thread to only be active in one
generic_make_request at a time.  If a subsequent (recursive) call is made,
the bio is linked into a per-thread list, and is handled when the active
call completes.

As the list of pending bios is per-thread, there are no locking issues to
worry about.

I say above that it is "safe to delay any call...".  There are, however,
some behaviours of a make_request_fn which would make it unsafe.  These
include any behaviour that assumes anything will have changed after a
recursive call to generic_make_request.

These could include:
 - waiting for that call to finish and call it's bi_end_io function.
   md use to sometimes do this (marking the superblock dirty before
   completing a write) but doesn't any more
 - inspecting the bio for fields that generic_make_request might
   change, such as bi_sector or bi_bdev.  It is hard to see a good
   reason for this, and I don't think anyone actually does it.
 - inspecing the queue to see if, e.g. it is 'full' yet.  Again, I
   think this is very unlikely to be useful, or to be done.

Signed-off-by: Neil Brown <neilb@suse.de>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: <dm-devel@redhat.com>
Alasdair G Kergon <agk@redhat.com> said:

 I can see nothing wrong with this in principle.

 For device-mapper at the moment though it's essential that, while the bio
 mappings may now get delayed, they still get processed in exactly
 the same order as they were passed to generic_make_request().

 My main concern is whether the timing changes implicit in this patch
 will make the rare data-corrupting races in the existing snapshot code
 more likely. (I'm working on a fix for these races, but the unfinished
 patch is already several hundred lines long.)

 It would be helpful if some people on this mailing list would test
 this patch in various scenarios and report back.

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
17 years ago[PATCH] Abnormal End of Processes
Steve Grubb [Thu, 19 Apr 2007 14:28:21 +0000 (10:28 -0400)]
[PATCH] Abnormal End of Processes

Hi,

I have been working on some code that detects abnormal events based on audit
system events. One kind of event that we currently have no visibility for is
when a program terminates due to segfault - which should never happen on a
production machine. And if it did, you'd want to investigate it. Attached is a
patch that collects these events and sends them into the audit system.

Signed-off-by: Steve Grubb <sgrubb@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[PATCH] match audit name data
Amy Griffis [Tue, 13 Feb 2007 19:15:22 +0000 (14:15 -0500)]
[PATCH] match audit name data

Make more effort to detect previously collected names, so we don't log
multiple PATH records for a single filesystem object. Add
audit_inc_name_count() to reduce duplicate code.

Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[PATCH] complete message queue auditing
Amy Griffis [Tue, 13 Feb 2007 19:15:01 +0000 (14:15 -0500)]
[PATCH] complete message queue auditing

Handle the edge cases for POSIX message queue auditing. Collect inode
info when opening an existing mq, and for send/receive operations. Remove
audit_inode_update() as it has really evolved into the equivalent of
audit_inode().

Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[PATCH] audit inode for all xattr syscalls
Amy Griffis [Tue, 13 Feb 2007 19:14:41 +0000 (14:14 -0500)]
[PATCH] audit inode for all xattr syscalls

Collect inode info for the remaining xattr syscalls that operate on a file
descriptor. These don't call a path_lookup variant, so they aren't covered by
the general audit hook.

Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[PATCH] initialize name osid
Amy Griffis [Tue, 13 Feb 2007 19:14:09 +0000 (14:14 -0500)]
[PATCH] initialize name osid

Audit contexts can be reused, so initialize a name's osid to the
default in audit_getname(). This ensures we don't log a bogus object
label when no inode data is collected for a name.

Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[PATCH] audit signal recipients
Amy Griffis [Thu, 29 Mar 2007 22:01:04 +0000 (18:01 -0400)]
[PATCH] audit signal recipients

When auditing syscalls that send signals, log the pid and security
context for each target process. Optimize the data collection by
adding a counter for signal-related rules, and avoiding allocating an
aux struct unless we have more than one target process. For process
groups, collect pid/context data in blocks of 16. Move the
audit_signal_info() hook up in check_kill_permission() so we audit
attempts where permission is denied.

Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[PATCH] add SIGNAL syscall class (v3)
Amy Griffis [Thu, 29 Mar 2007 22:00:37 +0000 (18:00 -0400)]
[PATCH] add SIGNAL syscall class (v3)

Add a syscall class for sending signals.

Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[PATCH] auditing ptrace
Al Viro [Tue, 20 Mar 2007 17:58:35 +0000 (13:58 -0400)]
[PATCH] auditing ptrace

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
17 years ago[NETFILTER]: xt_conntrack: add compat support
Patrick McHardy [Thu, 10 May 2007 21:17:58 +0000 (14:17 -0700)]
[NETFILTER]: xt_conntrack: add compat support

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: iptable_raw: ignore short packets sent by SOCK_RAW sockets
Patrick McHardy [Thu, 10 May 2007 21:17:36 +0000 (14:17 -0700)]
[NETFILTER]: iptable_raw: ignore short packets sent by SOCK_RAW sockets

iptables matches and targets expect packets to have at least a full
IP header and a valid header length. Ignore packets sent through
raw sockets for which this isn't true as in the other tables.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: iptable_{filter,mangle}: more descriptive "happy cracking" message
Patrick McHardy [Thu, 10 May 2007 21:16:56 +0000 (14:16 -0700)]
[NETFILTER]: iptable_{filter,mangle}: more descriptive "happy cracking" message

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: nf_nat: Clears helper private area when NATing
Yasuyuki Kozakai [Thu, 10 May 2007 21:16:24 +0000 (14:16 -0700)]
[NETFILTER]: nf_nat: Clears helper private area when NATing

Some helpers (eg. ftp) assume that private area in conntrack is
filled with zero. It should be cleared when helper is changed.

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: ctnetlink: clear helper area and handle unchanged helper
Yasuyuki Kozakai [Thu, 10 May 2007 21:15:58 +0000 (14:15 -0700)]
[NETFILTER]: ctnetlink: clear helper area and handle unchanged helper

This patch
- Clears private area for helper even if no helper is assigned to
  conntrack. It might be used by old helper.
- Unchanges if the same helper as the used one is specified.
- Does not find helper if no helper is specified. And it does not
  require private area for helper in that case.

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: nf_conntrack: Removes unused destroy operation of l3proto
Yasuyuki Kozakai [Thu, 10 May 2007 21:15:30 +0000 (14:15 -0700)]
[NETFILTER]: nf_conntrack: Removes unused destroy operation of l3proto

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: nf_conntrack: Removes duplicated declarations
Yasuyuki Kozakai [Thu, 10 May 2007 21:15:08 +0000 (14:15 -0700)]
[NETFILTER]: nf_conntrack: Removes duplicated declarations

These are also in include/net/netfilter/nf_conntrack_helper.h

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: nf_nat: remove unused argument of function allocating binding
Yasuyuki Kozakai [Thu, 10 May 2007 21:14:45 +0000 (14:14 -0700)]
[NETFILTER]: nf_nat: remove unused argument of function allocating binding

nf_nat_rule_find, alloc_null_binding and alloc_null_binding_confirmed
do not use the argument 'info', which is actually ct->nat.info.
If they are necessary to access it again, we can use the argument 'ct'
instead.

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NETFILTER]: Clean up table initialization
Patrick McHardy [Thu, 10 May 2007 21:14:16 +0000 (14:14 -0700)]
[NETFILTER]: Clean up table initialization

- move arp_tables initial table structure definitions to arp_tables.h
  similar to ip_tables and ip6_tables

- use C99 initializers

- use initializer macros where possible

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET_SCHED]: Avoid requeue warning on dev_deactivate
Herbert Xu [Thu, 10 May 2007 21:12:47 +0000 (14:12 -0700)]
[NET_SCHED]: Avoid requeue warning on dev_deactivate

When we relinquish queue_lock in qdisc_restart and then retake it for
requeueing, we might race against dev_deactivate and end up requeueing
onto noop_qdisc.  This causes a warning to be printed.

This patch fixes this by checking this before we requeue.  As an added
bonus, we can remove the same check in __qdisc_run which was added to
prevent dev->gso_skb from being requeued when we're shutting down.

Even though we've had to add a new conditional in its place, it's better
because it only happens on requeues rather than every single time that
qdisc_run is called.

For this to work we also need to move the clearing of gso_skb up in
dev_deactivate as now qdisc_restart can occur even after we wait for
__LINK_STATE_QDISC_RUNNING to clear (but it won't do anything as long
as the queue and gso_skb is already clear).

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET_SCHED]: Reread dev->qdisc for NETDEV_TX_OK
Herbert Xu [Thu, 10 May 2007 21:11:16 +0000 (14:11 -0700)]
[NET_SCHED]: Reread dev->qdisc for NETDEV_TX_OK

Now that we return the queue length after NETDEV_TX_OK we better
make sure that we have the right queue.  Otherwise we can cause a
stall after a really quick dev_deactive/dev_activate.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET_SCHED]: Rationalise return value of qdisc_restart
Herbert Xu [Thu, 10 May 2007 11:55:14 +0000 (04:55 -0700)]
[NET_SCHED]: Rationalise return value of qdisc_restart

The current return value scheme and associated comment was invented
back in the 20th century when we still had that tbusy flag.  Things
have changed quite a bit since then (even Tony Blair is moving on
now, not to mention the new French president).

All we need to indicate now is whether the caller should continue
processing the queue.  Therefore it's sufficient if we return 0 if
we want to stop and non-zero otherwise.

This is based on a patch by Krishna Kumar.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[NET]: Fix dev->qdisc race for NETDEV_TX_LOCKED case
Thomas Graf [Thu, 10 May 2007 11:02:41 +0000 (04:02 -0700)]
[NET]: Fix dev->qdisc race for NETDEV_TX_LOCKED case

When transmit fails with NETDEV_TX_LOCKED the skb is requeued
to dev->qdisc again. The dev->qdisc pointer is protected by
the queue lock which needs to be dropped when attempting to
transmit and acquired again before requeing. The problem is
that qdisc_restart() fetches the dev->qdisc pointer once and
stores it in the `q' variable which is invalidated when
dropping the queue_lock, therefore the variable needs to be
refreshed before requeueing.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[UDP]: Fix AF-specific references in AF-agnostic code.
David S. Miller [Wed, 9 May 2007 23:42:20 +0000 (16:42 -0700)]
[UDP]: Fix AF-specific references in AF-agnostic code.

__udp_lib_port_inuse() cannot make direct references to
inet_sk(sk)->rcv_saddr as that is ipv4 specific state and
this code is used by ipv6 too.

Use an operations vector to solve this, and this also paves
the way for ipv6 support for non-wild saddr hashing in UDP.

Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IrDA]: KingSun/DonShine USB IrDA dongle support.
Alex Villac�s Lasso [Wed, 9 May 2007 23:18:21 +0000 (16:18 -0700)]
[IrDA]: KingSun/DonShine USB IrDA dongle support.

This dongle does not follow the usb-irda specification, so it needs its
own special driver. In addition, it uses interrupt endpoints instead of
bulk ones as the rest of USB IrDA dongles supported by Linux (just to be
different?) and data reads need to be parsed to extract the valid bytes
before being unwrapped (details in the comment at the start of the
source). No speed commands have been discovered for this dongle, and I
suspect it does not have any at all.

On plugin, this dongle reports vendor and device IDs: 0x07c0:0x4200 .

The Windows driver that is used normally to control this dongle has a
filename of DSIR620.SYS .

Signed-off-by: Alex Villac�s Lasso <a_villacis@palosanto.com>
Signed-off-by: Samuel Ortiz <samuel@sortiz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPV6] ROUTE: Assign rt6i_idev for ip6_{prohibit,blk_hole}_entry.
YOSHIFUJI Hideaki [Wed, 9 May 2007 21:03:28 +0000 (14:03 -0700)]
[IPV6] ROUTE: Assign rt6i_idev for ip6_{prohibit,blk_hole}_entry.

I think this is less critical, but is also suitable for -stable
release.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[IPV6]: Do no rely on skb->dst before it is assigned.
YOSHIFUJI Hideaki [Wed, 9 May 2007 21:01:59 +0000 (14:01 -0700)]
[IPV6]: Do no rely on skb->dst before it is assigned.

Because skb->dst is assigned in ip6_route_input(), it is really
bad to use it in hop-by-hop option handler(s).

Closes: Bug #8450 (Eric Sesterhenn <snakebyte@gmx.de>)
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>