GitHub/MotorolaMobilityLLC/kernel-slsi.git
15 years agoepoll keyed wakeups: make sockets use keyed wakeups
Davide Libenzi [Tue, 31 Mar 2009 22:24:21 +0000 (15:24 -0700)]
epoll keyed wakeups: make sockets use keyed wakeups

Add support for event-aware wakeups to the sockets code.  Events are
delivered to the wakeup target, so that epoll can avoid spurious wakeups
for non-interesting events.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: introduce new *_poll() wakeup macros
Davide Libenzi [Tue, 31 Mar 2009 22:24:20 +0000 (15:24 -0700)]
epoll keyed wakeups: introduce new *_poll() wakeup macros

Introduce new wakeup macros that allow passing an event mask to the wakeup
targets.  They exactly mimic their non-_poll() counterpart, with the added
event mask passing capability.  I did add only the ones currently
requested, avoiding the _nr() and _all() for the moment.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: add __wake_up_locked_key() and __wake_up_sync_key()
Davide Libenzi [Tue, 31 Mar 2009 22:24:20 +0000 (15:24 -0700)]
epoll keyed wakeups: add __wake_up_locked_key() and __wake_up_sync_key()

This patchset introduces wakeup hints for some of the most popular (from
epoll POV) devices, so that epoll code can avoid spurious wakeups on its
waiters.

The problem with epoll is that the callback-based wakeups do not, ATM,
carry any information about the events the wakeup is related to.  So the
only choice epoll has (not being able to call f_op->poll() from inside the
callback), is to add the file* to a ready-list and resolve the real events
later on, at epoll_wait() (or its own f_op->poll()) time.  This can cause
spurious wakeups, since the wake_up() itself might be for an event the
caller is not interested into.

The rate of these spurious wakeup can be pretty high in case of many
network sockets being monitored.

By allowing devices to report the events the wakeups refer to (at least
the two major classes - POLLIN/POLLOUT), we are able to spare useless
wakeups by proper handling inside the epoll's poll callback.

Epoll will have in any case to call f_op->poll() on the file* later on,
since the change to be done in order to have the full event set sent via
wakeup, is too invasive for the way our f_op->poll() system works (the
full event set is calculated inside the poll function - there are too many
of them to even start thinking the change - also poll/select would need
change too).

Epoll is changed in a way that both devices which send event hints, and
the ones that don't, are correctly handled.  The former will gain some
efficiency though.

As a general rule for devices, would be to add an event mask by using
key-aware wakeup macros, when making up poll wait queues.  I tested it
(together with the epoll's poll fix patch Andrew has in -mm) and wakeups
for the supported devices are correctly filtered.

Test program available here:

http://www.xmailserver.org/epoll_test.c

This patch:

Nothing revolutionary here.  Just using the available "key" that our
wakeup core already support.  The __wake_up_locked_key() was no brainer,
since both __wake_up_locked() and __wake_up_locked_key() are thin wrappers
around __wake_up_common().

The __wake_up_sync() function had a body, so the choice was between
borrowing the body for __wake_up_sync_key() and calling it from
__wake_up_sync(), or make an inline and calling it from both.  I chose the
former since in most archs it all resolves to "mov $0, REG; jmp ADDR".

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoeventfd: improve support for semaphore-like behavior
Davide Libenzi [Tue, 31 Mar 2009 22:24:18 +0000 (15:24 -0700)]
eventfd: improve support for semaphore-like behavior

People started using eventfd in a semaphore-like way where before they
were using pipes.

That is, counter-based resource access.  Where a "wait()" returns
immediately by decrementing the counter by one, if counter is greater than
zero.  Otherwise will wait.  And where a "post(count)" will add count to
the counter releasing the appropriate amount of waiters.  If eventfd the
"post" (write) part is fine, while the "wait" (read) does not dequeue 1,
but the whole counter value.

The problem with eventfd is that a read() on the fd returns and wipes the
whole counter, making the use of it as semaphore a little bit more
cumbersome.  You can do a read() followed by a write() of COUNTER-1, but
IMO it's pretty easy and cheap to make this work w/out extra steps.  This
patch introduces a new eventfd flag that tells eventfd to only dequeue 1
from the counter, allowing simple read/write to make it behave like a
semaphore.  Simple test here:

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

To be back-compatible with earlier kernels, userspace applications should
probe for the availability of this feature via

#ifdef EFD_SEMAPHORE
fd = eventfd2 (CNT, EFD_SEMAPHORE);
if (fd == -1 && errno == EINVAL)
<fallback>
#else
<fallback>
#endif

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: <linux-api@vger.kernel.org>
Tested-by: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: use real type instead of void *
Tony Battersby [Tue, 31 Mar 2009 22:24:16 +0000 (15:24 -0700)]
epoll: use real type instead of void *

eventpoll.c uses void * in one place for no obvious reason; change it to
use the real type instead.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: clean up ep_modify
Tony Battersby [Tue, 31 Mar 2009 22:24:15 +0000 (15:24 -0700)]
epoll: clean up ep_modify

ep_modify() doesn't need to set event.data from within the ep->lock
spinlock as the comment suggests.  The only place event.data is used is
ep_send_events_proc(), and this is protected by ep->mtx instead of
ep->lock.  Also update the comment for mutex_lock() at the top of
ep_scan_ready_list(), which mentions epoll_ctl(EPOLL_CTL_DEL) but not
epoll_ctl(EPOLL_CTL_MOD).

ep_modify() can also use spin_lock_irq() instead of spin_lock_irqsave().

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: remove unnecessary xchg
Tony Battersby [Tue, 31 Mar 2009 22:24:15 +0000 (15:24 -0700)]
epoll: remove unnecessary xchg

xchg in ep_unregister_pollwait() is unnecessary because it is protected by
either epmutex or ep->mtx (the same protection as ep_remove()).

If xchg was necessary, it would be insufficient to protect against
problems: if multiple concurrent calls to ep_unregister_pollwait() were
possible then a second caller that returns without doing anything because
nwait == 0 could return before the waitqueues are removed by the first
caller, which looks like it could lead to problematic races with
ep_poll_callback().

So remove xchg and add comments about the locking.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: remember the event if epoll_wait returns -EFAULT
Tony Battersby [Tue, 31 Mar 2009 22:24:14 +0000 (15:24 -0700)]
epoll: remember the event if epoll_wait returns -EFAULT

If epoll_wait returns -EFAULT, the event that was being returned when the
fault was encountered will be forgotten.  This is not a big deal since
EFAULT will happen only if a buggy userspace program passes in a bad
address, in which case what happens later usually doesn't matter.
However, it is easy to remember the event for later, and this patch makes
a simple change to do that.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: don't use current in irq context
Tony Battersby [Tue, 31 Mar 2009 22:24:13 +0000 (15:24 -0700)]
epoll: don't use current in irq context

ep_call_nested() (formerly ep_poll_safewake()) uses "current" (without
dereferencing it) to detect callback recursion, but it may be called from
irq context where the use of current is generally discouraged.  It would
be better to use get_cpu() and put_cpu() to detect the callback recursion.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: remove debugging code
Davide Libenzi [Tue, 31 Mar 2009 22:24:12 +0000 (15:24 -0700)]
epoll: remove debugging code

Remove debugging code from epoll.  There's no need for it to be included
into mainline code.

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>
15 years agoepoll: fix epoll's own poll (update)
Davide Libenzi [Tue, 31 Mar 2009 22:24:11 +0000 (15:24 -0700)]
epoll: fix epoll's own poll (update)

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: fix epoll's own poll
Davide Libenzi [Tue, 31 Mar 2009 22:24:10 +0000 (15:24 -0700)]
epoll: fix epoll's own poll

Fix a bug inside the epoll's f_op->poll() code, that returns POLLIN even
though there are no actual ready monitored fds.  The bug shows up if you
add an epoll fd inside another fd container (poll, select, epoll).

The problem is that callback-based wake ups used by epoll does not carry
(patches will follow, to fix this) any information about the events that
actually happened.  So the callback code, since it can't call the file*
->poll() inside the callback, chains the file* into a ready-list.

So, suppose you added an fd with EPOLLOUT only, and some data shows up on
the fd, the file* mapped by the fd will be added into the ready-list (via
wakeup callback).  During normal epoll_wait() use, this condition is
sorted out at the time we're actually able to call the file*'s
f_op->poll().

Inside the old epoll's f_op->poll() though, only a quick check
!list_empty(ready-list) was performed, and this could have led to
reporting POLLIN even though no ready fds would show up at a following
epoll_wait().  In order to correctly report the ready status for an epoll
fd, the ready-list must be checked to see if any really available fd+event
would be ready in a following epoll_wait().

Operation (calling f_op->poll() from inside f_op->poll()) that, like wake
ups, must be handled with care because of the fact that epoll fds can be
added to other epoll fds.

Test code:

/*
 *  epoll_test by Davide Libenzi (Simple code to test epoll internals)
 *  Copyright (C) 2008  Davide Libenzi
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 *  Davide Libenzi <davidel@xmailserver.org>
 *
 */

#include <sys/types.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <signal.h>
#include <limits.h>
#include <poll.h>
#include <sys/epoll.h>
#include <sys/wait.h>

#define EPWAIT_TIMEO (1 * 1000)
#ifndef POLLRDHUP
#define POLLRDHUP 0x2000
#endif

#define EPOLL_MAX_CHAIN 100L

#define EPOLL_TF_LOOP (1 << 0)

struct epoll_test_cfg {
long size;
long flags;
};

static int xepoll_create(int n) {
int epfd;

if ((epfd = epoll_create(n)) == -1) {
perror("epoll_create");
exit(2);
}

return epfd;
}

static void xepoll_ctl(int epfd, int cmd, int fd, struct epoll_event *evt) {
if (epoll_ctl(epfd, cmd, fd, evt) < 0) {
perror("epoll_ctl");
exit(3);
}
}

static void xpipe(int *fds) {
if (pipe(fds)) {
perror("pipe");
exit(4);
}
}

static pid_t xfork(void) {
pid_t pid;

if ((pid = fork()) == (pid_t) -1) {
perror("pipe");
exit(5);
}

return pid;
}

static int run_forked_proc(int (*proc)(void *), void *data) {
int status;
pid_t pid;

if ((pid = xfork()) == 0)
exit((*proc)(data));
if (waitpid(pid, &status, 0) != pid) {
perror("waitpid");
return -1;
}

return WIFEXITED(status) ? WEXITSTATUS(status): -2;
}

static int check_events(int fd, int timeo) {
struct pollfd pfd;

fprintf(stdout, "Checking events for fd %d\n", fd);
memset(&pfd, 0, sizeof(pfd));
pfd.fd = fd;
pfd.events = POLLIN | POLLOUT;
if (poll(&pfd, 1, timeo) < 0) {
perror("poll()");
return 0;
}
if (pfd.revents & POLLIN)
fprintf(stdout, "\tPOLLIN\n");
if (pfd.revents & POLLOUT)
fprintf(stdout, "\tPOLLOUT\n");
if (pfd.revents & POLLERR)
fprintf(stdout, "\tPOLLERR\n");
if (pfd.revents & POLLHUP)
fprintf(stdout, "\tPOLLHUP\n");
if (pfd.revents & POLLRDHUP)
fprintf(stdout, "\tPOLLRDHUP\n");

return pfd.revents;
}

static int epoll_test_tty(void *data) {
int epfd, ifd = fileno(stdin), res;
struct epoll_event evt;

if (check_events(ifd, 0) != POLLOUT) {
fprintf(stderr, "Something is cooking on STDIN (%d)\n", ifd);
return 1;
}
epfd = xepoll_create(1);
fprintf(stdout, "Created epoll fd (%d)\n", epfd);
memset(&evt, 0, sizeof(evt));
evt.events = EPOLLIN;
xepoll_ctl(epfd, EPOLL_CTL_ADD, ifd, &evt);
if (check_events(epfd, 0) & POLLIN) {
res = epoll_wait(epfd, &evt, 1, 0);
if (res == 0) {
fprintf(stderr, "Epoll fd (%d) is ready when it shouldn't!\n",
epfd);
return 2;
}
}

return 0;
}

static int epoll_wakeup_chain(void *data) {
struct epoll_test_cfg *tcfg = data;
int i, res, epfd, bfd, nfd, pfds[2];
pid_t pid;
struct epoll_event evt;

memset(&evt, 0, sizeof(evt));
evt.events = EPOLLIN;

epfd = bfd = xepoll_create(1);

for (i = 0; i < tcfg->size; i++) {
nfd = xepoll_create(1);
xepoll_ctl(bfd, EPOLL_CTL_ADD, nfd, &evt);
bfd = nfd;
}
xpipe(pfds);
if (tcfg->flags & EPOLL_TF_LOOP)
{
xepoll_ctl(bfd, EPOLL_CTL_ADD, epfd, &evt);
/*
 * If we're testing for loop, we want that the wakeup
 * triggered by the write to the pipe done in the child
 * process, triggers a fake event. So we add the pipe
 * read size with EPOLLOUT events. This will trigger
 * an addition to the ready-list, but no real events
 * will be there. The the epoll kernel code will proceed
 * in calling f_op->poll() of the epfd, triggering the
 * loop we want to test.
 */
evt.events = EPOLLOUT;
}
xepoll_ctl(bfd, EPOLL_CTL_ADD, pfds[0], &evt);

/*
 * The pipe write must come after the poll(2) call inside
 * check_events(). This tests the nested wakeup code in
 * fs/eventpoll.c:ep_poll_safewake()
 * By having the check_events() (hence poll(2)) happens first,
 * we have poll wait queue filled up, and the write(2) in the
 * child will trigger the wakeup chain.
 */
if ((pid = xfork()) == 0) {
sleep(1);
write(pfds[1], "w", 1);
exit(0);
}

res = check_events(epfd, 2000) & POLLIN;

if (waitpid(pid, NULL, 0) != pid) {
perror("waitpid");
return -1;
}

return res;
}

static int epoll_poll_chain(void *data) {
struct epoll_test_cfg *tcfg = data;
int i, res, epfd, bfd, nfd, pfds[2];
pid_t pid;
struct epoll_event evt;

memset(&evt, 0, sizeof(evt));
evt.events = EPOLLIN;

epfd = bfd = xepoll_create(1);

for (i = 0; i < tcfg->size; i++) {
nfd = xepoll_create(1);
xepoll_ctl(bfd, EPOLL_CTL_ADD, nfd, &evt);
bfd = nfd;
}
xpipe(pfds);
if (tcfg->flags & EPOLL_TF_LOOP)
{
xepoll_ctl(bfd, EPOLL_CTL_ADD, epfd, &evt);
/*
 * If we're testing for loop, we want that the wakeup
 * triggered by the write to the pipe done in the child
 * process, triggers a fake event. So we add the pipe
 * read size with EPOLLOUT events. This will trigger
 * an addition to the ready-list, but no real events
 * will be there. The the epoll kernel code will proceed
 * in calling f_op->poll() of the epfd, triggering the
 * loop we want to test.
 */
evt.events = EPOLLOUT;
}
xepoll_ctl(bfd, EPOLL_CTL_ADD, pfds[0], &evt);

/*
 * The pipe write mush come before the poll(2) call inside
 * check_events(). This tests the nested f_op->poll calls code in
 * fs/eventpoll.c:ep_eventpoll_poll()
 * By having the pipe write(2) happen first, we make the kernel
 * epoll code to load the ready lists, and the following poll(2)
 * done inside check_events() will test nested poll code in
 * ep_eventpoll_poll().
 */
if ((pid = xfork()) == 0) {
write(pfds[1], "w", 1);
exit(0);
}
sleep(1);
res = check_events(epfd, 1000) & POLLIN;

if (waitpid(pid, NULL, 0) != pid) {
perror("waitpid");
return -1;
}

return res;
}

int main(int ac, char **av) {
int error;
struct epoll_test_cfg tcfg;

fprintf(stdout, "\n********** Testing TTY events\n");
error = run_forked_proc(epoll_test_tty, NULL);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing short wakeup chain\n");
error = run_forked_proc(epoll_wakeup_chain, &tcfg);
fprintf(stdout, error == POLLIN ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = EPOLL_MAX_CHAIN;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing long wakeup chain (HOLD ON)\n");
error = run_forked_proc(epoll_wakeup_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing short poll chain\n");
error = run_forked_proc(epoll_poll_chain, &tcfg);
fprintf(stdout, error == POLLIN ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = EPOLL_MAX_CHAIN;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing long poll chain (HOLD ON)\n");
error = run_forked_proc(epoll_poll_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = EPOLL_TF_LOOP;
fprintf(stdout, "\n********** Testing loopy wakeup chain (HOLD ON)\n");
error = run_forked_proc(epoll_wakeup_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = EPOLL_TF_LOOP;
fprintf(stdout, "\n********** Testing loopy poll chain (HOLD ON)\n");
error = run_forked_proc(epoll_poll_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

return 0;
}

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrivers/misc/isl29003.c: driver for the ISL29003 ambient light sensor
Daniel Mack [Tue, 31 Mar 2009 22:23:53 +0000 (15:23 -0700)]
drivers/misc/isl29003.c: driver for the ISL29003 ambient light sensor

Add a driver for Intersil's ISL29003 ambient light sensor device plus some
documentation.  Inspired by tsl2550.c, a driver for a similar device.

It is put in drivers/misc for now until the industrial I/O framework gets
merged.

Signed-off-by: Daniel Mack <daniel@caiaq.de>
Acked-by: Jonathan Cameron <jic23@cam.ac.uk>
Cc: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohpilo: reduce frequency of IO operations
David Altobelli [Tue, 31 Mar 2009 22:23:53 +0000 (15:23 -0700)]
hpilo: reduce frequency of IO operations

Change hpilo open and close logic to spin for 10usec between checking device,
rather than every usec.

Because the loop is coded to take up to 10ms, it seemed prudent to
increase the interval between polling the device, to reduce the load on
the system and allow more other work to happen.

Signed-off-by: David Altobelli <david.altobelli@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agontfs: remove private wrapper of endian helpers
Harvey Harrison [Tue, 31 Mar 2009 22:23:52 +0000 (15:23 -0700)]
ntfs: remove private wrapper of endian helpers

The base versions handle constant folding now and are shorter than these
private wrappers, use them directly.

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Cc: Anton Altaparmakov <aia21@cantab.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agointroduce pr_cont() macro
Cyrill Gorcunov [Tue, 31 Mar 2009 22:23:51 +0000 (15:23 -0700)]
introduce pr_cont() macro

We cover all log-levels by pr_...  macros except KERN_CONT one.  Add it
for convenience.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoinit/main.c: fix sparse warnings: context imbalance
Hannes Eder [Tue, 31 Mar 2009 22:23:50 +0000 (15:23 -0700)]
init/main.c: fix sparse warnings: context imbalance

Impact: Attribute function 'init_post' with __releases(...).

Fix these sparse warnings:
  init/main.c:805:21: warning: context imbalance in 'init_post' - unexpected unlock
  init/main.c:899:9: warning: context imbalance in 'kernel_init' - wrong count at exit

Signed-off-by: Hannes Eder <hannes@hanneseder.net>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agobcm47xx: fix GPIO API return codes
Michael Buesch [Tue, 31 Mar 2009 22:23:49 +0000 (15:23 -0700)]
bcm47xx: fix GPIO API return codes

The GPIO API is supposed to return 0 or a negative error code,
but the SSB GPIO functions return the bitmask of the GPIO register.
Fix this by ignoring the bitmask and always returning 0. The SSB GPIO functions can't fail.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Cc: Ralf Baechle <ralf@linux-mips.org>
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>
15 years agoauxdisplay: remove PARPORT dependency
H Hartley Sweeten [Tue, 31 Mar 2009 22:23:48 +0000 (15:23 -0700)]
auxdisplay: remove PARPORT dependency

Remove PARPORT dependency for Auxiliary Display support.

This is not needed since the dependency for the KS0108 driver is
PARPORT_PC.

Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Miguel Ojeda Sandonis <miguel.ojeda.sandonis@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoremove unused include/asm-generic/dma-mapping.h
FUJITA Tomonori [Tue, 31 Mar 2009 22:23:47 +0000 (15:23 -0700)]
remove unused include/asm-generic/dma-mapping.h

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofilesystem freeze: allow SysRq emergency thaw to thaw frozen filesystems
Eric Sandeen [Tue, 31 Mar 2009 22:23:46 +0000 (15:23 -0700)]
filesystem freeze: allow SysRq emergency thaw to thaw frozen filesystems

Now that the filesystem freeze operation has been elevated to the VFS, and
is just an ioctl away, some sort of safety net for unintentionally frozen
root filesystems may be in order.

The timeout thaw originally proposed did not get merged, but perhaps
something like this would be useful in emergencies.

For example, freeze /path/to/mountpoint may freeze your root filesystem if
you forgot that you had that unmounted.

I chose 'j' as the last remaining character other than 'h' which is sort
of reserved for help (because help is generated on any unknown character).

I've tested this on a non-root fs with multiple (nested) freezers, as well
as on a system rendered unresponsive due to a frozen root fs.

[randy.dunlap@oracle.com: emergency thaw only if CONFIG_BLOCK enabled]
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Cc: Takashi Sato <t-sato@yk.jp.nec.com>
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>
15 years agolib/rbtree.c: optimize rb_erase()
Wolfram Strepp [Tue, 31 Mar 2009 22:23:45 +0000 (15:23 -0700)]
lib/rbtree.c: optimize rb_erase()

Tfour 4 redundant if-conditions in function __rb_erase_color() in
lib/rbtree.c are removed.

In pseudo-source-code, the structure of the code is as follows:

if ((!A || B) && (!C || D)) {
.
.
.
} else {
if (!C || D) {//if this is true, it implies: (A == true) && (B == false)
if (A) {//hence this always evaluates to 'true'...
.
}
.
//at this point, C always becomes true, because of:
__rb_rotate_right/left();
//and:
other = parent->rb_right/left;
}
.
.
if (C) {//...and this too !
.
}
}

Signed-off-by: Wolfram Strepp <wstrepp@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoloop: add ioctl to resize a loop device
J. R. Okajima [Tue, 31 Mar 2009 22:23:43 +0000 (15:23 -0700)]
loop: add ioctl to resize a loop device

Add the ability to 'resize' the loop device on the fly.

One practical application is a loop file with XFS filesystem, already
mounted: You can easily enlarge the file (append some bytes) and then call
ioctl(fd, LOOP_SET_CAPACITY, new); The loop driver will learn about the
new size and you can use xfs_growfs later on, which will allow you to use
full capacity of the loop file without the need to unmount.

Test app:

#include <linux/fs.h>
#include <linux/loop.h>
#include <sys/ioctl.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <assert.h>
#include <errno.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

#define _GNU_SOURCE
#include <getopt.h>

char *me;

void usage(FILE *f)
{
fprintf(f, "%s [options] loop_dev [backend_file]\n"
"-s, --set new_size_in_bytes\n"
"\twhen backend_file is given, "
"it will be expanded too while keeping the original contents\n",
me);
}

struct option opts[] = {
{
.name = "set",
.has_arg = 1,
.flag = NULL,
.val = 's'
},
{
.name = "help",
.has_arg = 0,
.flag = NULL,
.val = 'h'
}
};

void err_size(char *name, __u64 old)
{
fprintf(stderr, "size must be larger than current %s (%llu)\n",
name, old);
}

int main(int argc, char *argv[])
{
int fd, err, c, i, bfd;
ssize_t ssz;
size_t sz;
__u64 old, new, append;
char a[BUFSIZ];
struct stat st;
FILE *out;
char *backend, *dev;

err = EINVAL;
out = stderr;
me = argv[0];
new = 0;
while ((c = getopt_long(argc, argv, "s:h", opts, &i)) != -1) {
switch (c) {
case 's':
errno = 0;
new = strtoull(optarg, NULL, 0);
if (errno) {
err = errno;
perror(argv[i]);
goto out;
}
break;

case 'h':
err = 0;
out = stdout;
goto err;

default:
perror(argv[i]);
goto err;
}
}

if (optind < argc)
dev = argv[optind++];
else
goto err;

fd = open(dev, O_RDONLY);
if (fd < 0) {
err = errno;
perror(dev);
goto out;
}

err = ioctl(fd, BLKGETSIZE64, &old);
if (err) {
err = errno;
perror("ioctl BLKGETSIZE64");
goto out;
}

if (!new) {
printf("%llu\n", old);
goto out;
}

if (new < old) {
err = EINVAL;
err_size(dev, old);
goto out;
}

if (optind < argc) {
backend = argv[optind++];
bfd = open(backend, O_WRONLY|O_APPEND);
if (bfd < 0) {
err = errno;
perror(backend);
goto out;
}
err = fstat(bfd, &st);
if (err) {
err = errno;
perror(backend);
goto out;
}
if (new < st.st_size) {
err = EINVAL;
err_size(backend, st.st_size);
goto out;
}
append = new - st.st_size;
sz = sizeof(a);
while (append > 0) {
if (append < sz)
sz = append;
ssz = write(bfd, a, sz);
if (ssz != sz) {
err = errno;
perror(backend);
goto out;
}
append -= sz;
}
err = fsync(bfd);
if (err) {
err = errno;
perror(backend);
goto out;
}
}

err = ioctl(fd, LOOP_SET_CAPACITY, new);
if (err) {
err = errno;
perror("ioctl LOOP_SET_CAPACITY");
}
goto out;

 err:
usage(out);
 out:
return err;
}

Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>
Signed-off-by: Tomas Matejicek <tomas@slax.org>
Cc: <util-linux-ng@vger.kernel.org>
Cc: Karel Zak <kzak@redhat.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Akinobu Mita <akinobu.mita@gmail.com>
Cc: <linux-api@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agouml: remove useless comments
WANG Cong [Tue, 31 Mar 2009 22:23:41 +0000 (15:23 -0700)]
uml: remove useless comments

These comments are useless now, remove them.

Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agouml: improve error messages
WANG Cong [Tue, 31 Mar 2009 22:23:41 +0000 (15:23 -0700)]
uml: improve error messages

These error messages are from check_sysemu(), not check_ptrace().

Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agouml: don't use a too long string literal
WANG Cong [Tue, 31 Mar 2009 22:23:40 +0000 (15:23 -0700)]
uml: don't use a too long string literal

uml uses a concatenated string literal to store the contents of .config,
but .config file content is varaible, it can be very long.

Use an array of string literals instead.

Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoubd: stop defintining MAJOR_NR
Christoph Hellwig [Tue, 31 Mar 2009 22:23:39 +0000 (15:23 -0700)]
ubd: stop defintining MAJOR_NR

MAJOR_NR isn't needed anymore since very early 2.5 kernels.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agopm: cleanup includes
Magnus Damm [Tue, 31 Mar 2009 22:23:38 +0000 (15:23 -0700)]
pm: cleanup includes

Remove unused/duplicate cruft from asm/suspend.h:

 - x86_32: remove unused acpi code
 - powerpc: remove duplicate prototypes, see linux/suspend.h

Signed-off-by: Magnus Damm <damm@igel.co.jp>
Cc: Paul Mundt <lethal@linux-sh.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agopm: rework includes, remove arch ifdefs
Magnus Damm [Tue, 31 Mar 2009 22:23:37 +0000 (15:23 -0700)]
pm: rework includes, remove arch ifdefs

Make the following header file changes:

 - remove arch ifdefs and asm/suspend.h from linux/suspend.h
 - add asm/suspend.h to disk.c (for arch_prepare_suspend())
 - add linux/io.h to swsusp.c (for ioremap())
 - x86 32/64 bit compile fixes

Signed-off-by: Magnus Damm <damm@igel.co.jp>
Cc: Paul Mundt <lethal@linux-sh.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoalpha: convert u64 to unsigned long long
Randy Dunlap [Tue, 31 Mar 2009 22:23:36 +0000 (15:23 -0700)]
alpha: convert u64 to unsigned long long

Convert alpha architecture to use u64 as unsigned long long.  This is
being done so that (a) all arches use u64 as unsigned long long and (b)
printk of a u64 as %ll[ux] will not generate format warnings by gcc.

The only gcc cross-compiler that I have is 4.0.2, which generates errors
about miscompiling __weak references, so I have commented out that line in
compiler-gcc4.h so that most of these compile, but more builds and real
machine testing would be Real Good.

[akpm@linux-foundation.org: fix warning]
[akpm@linux-foundation.org: fix build]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
From: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoalpha: xchg/cmpxchg cleanup and fixes
Ivan Kokshaysky [Tue, 31 Mar 2009 22:23:35 +0000 (15:23 -0700)]
alpha: xchg/cmpxchg cleanup and fixes

- "_local" versions of xchg/cmpxchg functions duplicate code
  of non-local ones (quite a few pages of assembler), except
  memory barriers. We can generate these two variants from a
  single header file using simple macros;

- convert xchg macro back to inline function using always_inline
  attribute;

- use proper argument types for cmpxchg_u8/u16 functions
  to fix a problem with negative arguments.

Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMAINTAINERS: add the missing linux alpha port mailling list
Cheng Renquan [Tue, 31 Mar 2009 22:23:35 +0000 (15:23 -0700)]
MAINTAINERS: add the missing linux alpha port mailling list

Signed-off-by: Cheng Renquan <crquan@gmail.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoalpha: fix macros
Roel Kluin [Tue, 31 Mar 2009 22:23:34 +0000 (15:23 -0700)]
alpha: fix macros

When this macros isn't called with 'fixup', e.g.  with foo this will
incorectly expand to foo->foo.bits.errreg

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoshmem: writepage directly to swap
Hugh Dickins [Tue, 31 Mar 2009 22:23:33 +0000 (15:23 -0700)]
shmem: writepage directly to swap

Synopsis: if shmem_writepage calls swap_writepage directly, most shmem
swap loads benefit, and a catastrophic interaction between SLUB and some
flash storage is avoided.

shmem_writepage() has always been peculiar in making no attempt to write:
it has just transferred a shmem page from file cache to swap cache, then
let that page make its way around the LRU again before being written and
freed.

The idea was that people use tmpfs because they want those pages to stay
in RAM; so although we give it an overflow to swap, we should resist
writing too soon, giving those pages a second chance before they can be
reclaimed.

That was always questionable, and I've toyed with this patch for years;
but never had a clear justification to depart from the original design.

It became more questionable in 2.6.28, when the split LRU patches classed
shmem and tmpfs pages as SwapBacked rather than as file_cache: that in
itself gives them more resistance to reclaim than normal file pages.  I
prepared this patch for 2.6.29, but the merge window arrived before I'd
completed gathering statistics to justify sending it in.

Then while comparing SLQB against SLUB, running SLUB on a laptop I'd
habitually used with SLAB, I found SLUB to run my tmpfs kbuild swapping
tests five times slower than SLAB or SLQB - other machines slower too, but
nowhere near so bad.  Simpler "cp -a" swapping tests showed the same.

slub_max_order=0 brings sanity to all, but heavy swapping is too far from
normal to justify such a tuning.  The crucial factor on that laptop turns
out to be that I'm using an SD card for swap.  What happens is this:

By default, SLUB uses order-2 pages for shmem_inode_cache (and many other
fs inodes), so creating tmpfs files under memory pressure brings lumpy
reclaim into play.  One subpage of the order is chosen from the bottom of
the LRU as usual, then the other three picked out from their random
positions on the LRUs.

In a tmpfs load, many of these pages will be ones which already passed
through shmem_writepage, so already have swap allocated.  And though their
offsets on swap were probably allocated sequentially, now that the pages
are picked off at random, their swap offsets are scattered.

But the flash storage on the SD card is very sensitive to having its
writes merged: once swap is written at scattered offsets, performance
falls apart.  Rotating disk seeks increase too, but less disastrously.

So: stop giving shmem/tmpfs pages a second pass around the LRU, write them
out to swap as soon as their swap has been allocated.

It's surely possible to devise an artificial load which runs faster the
old way, one whose sizing is such that the tmpfs pages on their second
pass are the ones that are wanted again, and other pages not.

But I've not yet found such a load: on all machines, under the loads I've
tried, immediate swap_writepage speeds up shmem swapping: especially when
using the SLUB allocator (and more effectively than slub_max_order=0), but
also with the others; and it also reduces the variance between runs.  How
much faster varies widely: a factor of five is rare, 5% is common.

One load which might have suffered: imagine a swapping shmem load in a
limited mem_cgroup on a machine with plenty of memory.  Before 2.6.29 the
swapcache was not charged, and such a load would have run quickest with
the shmem swapcache never written to swap.  But now swapcache is charged,
so even this load benefits from shmem_writepage directly to swap.

Apologies for the #ifndef CONFIG_SWAP swap_writepage() stub in swap.h:
it's silly because that will never get called; but refactoring shmem.c
sensibly according to CONFIG_SWAP will be a separate task.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovmscan: fix it to take care of nodemask
KAMEZAWA Hiroyuki [Tue, 31 Mar 2009 22:23:31 +0000 (15:23 -0700)]
vmscan: fix it to take care of nodemask

try_to_free_pages() is used for the direct reclaim of up to
SWAP_CLUSTER_MAX pages when watermarks are low.  The caller to
alloc_pages_nodemask() can specify a nodemask of nodes that are allowed to
be used but this is not passed to try_to_free_pages().  This can lead to
unnecessary reclaim of pages that are unusable by the caller and int the
worst case lead to allocation failure as progress was not been make where
it is needed.

This patch passes the nodemask used for alloc_pages_nodemask() to
try_to_free_pages().

Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
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>
15 years agoramfs-nommu: use generic lru cache
Johannes Weiner [Tue, 31 Mar 2009 22:23:29 +0000 (15:23 -0700)]
ramfs-nommu: use generic lru cache

Instead of open-coding the lru-list-add pagevec batching when expanding a
file mapping from zero, defer to the appropriate page cache function that
also takes care of adding the page to the lru list.

This is cleaner, saves code and reduces the stack footprint by 16 words
worth of pagevec.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Howells <dhowells@redhat.com>
Cc: Nick Piggin <npiggin@suse.de>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.com>
Cc: MinChan Kim <minchan.kim@gmail.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovmscan: print shrink_slab symbol name on negative shrinker objects
David Rientjes [Tue, 31 Mar 2009 22:23:29 +0000 (15:23 -0700)]
vmscan: print shrink_slab symbol name on negative shrinker objects

When a shrinker has a negative number of objects to delete, the symbol
name of the shrinker should be printed, not shrink_slab.  This also makes
the error message slightly more informative.

Cc: Ingo Molnar <mingo@elte.hu>
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>
15 years agonommu: make CONFIG_UNEVICTABLE_LRU available when CONFIG_MMU=n
David Howells [Tue, 31 Mar 2009 22:23:28 +0000 (15:23 -0700)]
nommu: make CONFIG_UNEVICTABLE_LRU available when CONFIG_MMU=n

Make CONFIG_UNEVICTABLE_LRU available when CONFIG_MMU=n.  There's no logical
reason it shouldn't be available, and it can be used for ramfs.

Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Greg Ungerer <gerg@snapgear.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Enrik Berkhan <Enrik.Berkhan@ge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agonommu: there is no mlock() for NOMMU, so don't provide the bits
David Howells [Tue, 31 Mar 2009 22:23:26 +0000 (15:23 -0700)]
nommu: there is no mlock() for NOMMU, so don't provide the bits

The mlock() facility does not exist for NOMMU since all mappings are
effectively locked anyway, so we don't make the bits available when
they're not useful.

Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Greg Ungerer <gerg@snapgear.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Enrik Berkhan <Enrik.Berkhan@ge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: use debug_kmap_atomic
Akinobu Mita [Tue, 31 Mar 2009 22:23:25 +0000 (15:23 -0700)]
mm: use debug_kmap_atomic

Use debug_kmap_atomic in kmap_atomic, kmap_atomic_pfn, and
iomap_atomic_prot_pfn.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: introduce debug_kmap_atomic
Akinobu Mita [Tue, 31 Mar 2009 22:23:25 +0000 (15:23 -0700)]
mm: introduce debug_kmap_atomic

x86 has debug_kmap_atomic_prot() which is error checking function for
kmap_atomic.  It is usefull for the other architectures, although it needs
CONFIG_TRACE_IRQFLAGS_SUPPORT.

This patch exposes it to the other architectures.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: page_mkwrite change prototype to match fault: fix sysfs
Hugh Dickins [Tue, 31 Mar 2009 22:23:24 +0000 (15:23 -0700)]
mm: page_mkwrite change prototype to match fault: fix sysfs

Fix warnings and return values in sysfs bin_page_mkwrite(), fixing
fs/sysfs/bin.c: In function `bin_page_mkwrite':
fs/sysfs/bin.c:250: warning: passing argument 2 of `bb->vm_ops->page_mkwrite' from incompatible pointer type
fs/sysfs/bin.c: At top level:
fs/sysfs/bin.c:280: warning: initialization from incompatible pointer type

Expects to have my [PATCH next] sysfs: fix some bin_vm_ops errors

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: "Eric W. Biederman" <ebiederm@aristanetworks.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofs: fix page_mkwrite error cases in core code and btrfs
Nick Piggin [Tue, 31 Mar 2009 22:23:23 +0000 (15:23 -0700)]
fs: fix page_mkwrite error cases in core code and btrfs

page_mkwrite is called with neither the page lock nor the ptl held.  This
means a page can be concurrently truncated or invalidated out from
underneath it.  Callers are supposed to prevent truncate races themselves,
however previously the only thing they can do in case they hit one is to
raise a SIGBUS.  A sigbus is wrong for the case that the page has been
invalidated or truncated within i_size (eg.  hole punched).  Callers may
also have to perform memory allocations in this path, where again, SIGBUS
would be wrong.

The previous patch ("mm: page_mkwrite change prototype to match fault")
made it possible to properly specify errors.  Convert the generic buffer.c
code and btrfs to return sane error values (in the case of page removed
from pagecache, VM_FAULT_NOPAGE will cause the fault handler to exit
without doing anything, and the fault will be retried properly).

This fixes core code, and converts btrfs as a template/example.  All other
filesystems defining their own page_mkwrite should be fixed in a similar
manner.

Acked-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: page_mkwrite change prototype to match fault
Nick Piggin [Tue, 31 Mar 2009 22:23:21 +0000 (15:23 -0700)]
mm: page_mkwrite change prototype to match fault

Change the page_mkwrite prototype to take a struct vm_fault, and return
VM_FAULT_xxx flags.  There should be no functional change.

This makes it possible to return much more detailed error information to
the VM (and also can provide more information eg.  virtual_address to the
driver, which might be important in some special cases).

This is required for a subsequent fix.  And will also make it easier to
merge page_mkwrite() with fault() in future.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <joel.becker@oracle.com>
Cc: Artem Bityutskiy <dedekind@infradead.org>
Cc: Felix Blyakher <felixb@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: enable hashdist by default on 64bit NUMA
Anton Blanchard [Tue, 31 Mar 2009 22:23:19 +0000 (15:23 -0700)]
mm: enable hashdist by default on 64bit NUMA

On PowerPC we allocate large boot time hashes on node 0.  This leads to an
imbalance in the free memory, for example on a 64GB box (4 x 16GB nodes):

Free memory:
Node 0: 97.03%
Node 1: 98.54%
Node 2: 98.42%
Node 3: 98.53%

If we switch to using vmalloc (like ia64 and x86-64) things are more
balanced:

Free memory:
Node 0: 97.53%
Node 1: 98.35%
Node 2: 98.33%
Node 3: 98.33%

For many HPC applications we are limited by the free available memory on
the smallest node, so even though the same amount of memory is used the
better balancing helps.

Since all 64bit NUMA capable architectures should have sufficient vmalloc
space, it makes sense to enable it via CONFIG_64BIT.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: fix proc_dointvec_userhz_jiffies "breakage"
Alexey Dobriyan [Tue, 31 Mar 2009 22:23:18 +0000 (15:23 -0700)]
mm: fix proc_dointvec_userhz_jiffies "breakage"

Addresses http://bugzilla.kernel.org/show_bug.cgi?id=9838

On i386, HZ=1000, jiffies_to_clock_t() converts time in a somewhat strange
way from the user's point of view:

# echo 500 >/proc/sys/vm/dirty_writeback_centisecs
# cat /proc/sys/vm/dirty_writeback_centisecs
499

So, we have 5000 jiffies converted to only 499 clock ticks and reported
back.

TICK_NSEC = 999848
ACTHZ = 256039

Keeping in-kernel variable in units passed from userspace will fix issue
of course, but this probably won't be right for every sysctl.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agogeneric debug pagealloc
Akinobu Mita [Tue, 31 Mar 2009 22:23:17 +0000 (15:23 -0700)]
generic debug pagealloc

CONFIG_DEBUG_PAGEALLOC is now supported by x86, powerpc, sparc64, and
s390.  This patch implements it for the rest of the architectures by
filling the pages with poison byte patterns after free_pages() and
verifying the poison patterns before alloc_pages().

This generic one cannot detect invalid page accesses immediately but
invalid read access may cause invalid dereference by poisoned memory and
invalid write access can be detected after a long delay.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomemdup_user(): introduce
Li Zefan [Tue, 31 Mar 2009 22:23:16 +0000 (15:23 -0700)]
memdup_user(): introduce

I notice there are many places doing copy_from_user() which follows
kmalloc():

        dst = kmalloc(len, GFP_KERNEL);
        if (!dst)
                return -ENOMEM;
        if (copy_from_user(dst, src, len)) {
kfree(dst);
return -EFAULT
}

memdup_user() is a wrapper of the above code.  With this new function, we
don't have to write 'len' twice, which can lead to typos/mistakes.  It
also produces smaller code and kernel text.

A quick grep shows 250+ places where memdup_user() *may* be used.  I'll
prepare a patchset to do this conversion.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Americo Wang <xiyou.wangcong@gmail.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohugetlb: chg cannot become less than 0
Roel Kluin [Tue, 31 Mar 2009 22:23:15 +0000 (15:23 -0700)]
hugetlb: chg cannot become less than 0

chg is unsigned, so it cannot be less than 0.

Also, since region_chg returns long, let vma_needs_reservation() forward
this to alloc_huge_page().  Store it as long as well.  all callers cast it
to long anyway.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Adam Litke <agl@us.ibm.com>
Cc: Johannes Weiner <hannes@saeurebad.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: remove pagevec_swap_free()
KOSAKI Motohiro [Tue, 31 Mar 2009 22:23:14 +0000 (15:23 -0700)]
mm: remove pagevec_swap_free()

pagevec_swap_free() is now unused.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: don't free swap slots on page deactivation
Johannes Weiner [Tue, 31 Mar 2009 22:23:13 +0000 (15:23 -0700)]
mm: don't free swap slots on page deactivation

The pagevec_swap_free() at the end of shrink_active_list() was introduced
in 68a22394 "vmscan: free swap space on swap-in/activation" when
shrink_active_list() was still rotating referenced active pages.

In 7e9cd48 "vmscan: fix pagecache reclaim referenced bit check" this was
changed, the rotating removed but the pagevec_swap_free() after the
rotation loop was forgotten, applying now to the pagevec of the
deactivation loop instead.

Now swap space is freed for deactivated pages.  And only for those that
happen to be on the pagevec after the deactivation loop.

Complete 7e9cd48 and remove the rest of the swap freeing.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: move pagevec stripping to save unlock-relock
Johannes Weiner [Tue, 31 Mar 2009 22:23:12 +0000 (15:23 -0700)]
mm: move pagevec stripping to save unlock-relock

In shrink_active_list() after the deactivation loop, we strip buffer heads
from the potentially remaining pages in the pagevec.

Currently, this drops the zone's lru lock for stripping, only to reacquire
it again afterwards to update statistics.

It is not necessary to strip the pages before updating the stats, so move
the whole thing out of the protected region and save the extra locking.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: MinChan Kim <minchan.kim@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: reintroduce and deprecate rlimit based access for SHM_HUGETLB
Ravikiran G Thirumalai [Tue, 31 Mar 2009 22:21:26 +0000 (15:21 -0700)]
mm: reintroduce and deprecate rlimit based access for SHM_HUGETLB

Allow non root users with sufficient mlock rlimits to be able to allocate
hugetlb backed shm for now.  Deprecate this though.  This is being
deprecated because the mlock based rlimit checks for SHM_HUGETLB is not
consistent with mmap based huge page allocations.

Signed-off-by: Ravikiran Thirumalai <kiran@scalex86.org>
Reviewed-by: Mel Gorman <mel@csn.ul.ie>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Adam Litke <agl@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: fix SHM_HUGETLB to work with users in hugetlb_shm_group
Ravikiran G Thirumalai [Tue, 31 Mar 2009 22:19:40 +0000 (15:19 -0700)]
mm: fix SHM_HUGETLB to work with users in hugetlb_shm_group

Fix hugetlb subsystem so that non root users belonging to
hugetlb_shm_group can actually allocate hugetlb backed shm.

Currently non root users cannot even map one large page using SHM_HUGETLB
when they belong to the gid in /proc/sys/vm/hugetlb_shm_group.  This is
because allocation size is verified against RLIMIT_MEMLOCK resource limit
even if the user belongs to hugetlb_shm_group.

This patch
1. Fixes hugetlb subsystem so that users with CAP_IPC_LOCK and users
   belonging to hugetlb_shm_group don't need to be restricted with
   RLIMIT_MEMLOCK resource limits
2. This patch also disables mlock based rlimit checking (which will
   be reinstated and marked deprecated in a subsequent patch).

Signed-off-by: Ravikiran Thirumalai <kiran@scalex86.org>
Reviewed-by: Mel Gorman <mel@csn.ul.ie>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Adam Litke <agl@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovfs: add/use account_page_dirtied()
Edward Shishkin [Tue, 31 Mar 2009 22:19:39 +0000 (15:19 -0700)]
vfs: add/use account_page_dirtied()

Add a helper function account_page_dirtied().  Use that from two
callsites.  reiser4 adds a function which adds a third callsite.

Signed-off-by: Edward Shishkin<edward.shishkin@gmail.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovmscan: respect higher order in zone_reclaim()
Johannes Weiner [Tue, 31 Mar 2009 22:19:38 +0000 (15:19 -0700)]
vmscan: respect higher order in zone_reclaim()

During page allocation, there are two stages of direct reclaim that are
applied to each zone in the preferred list.  The first stage using
zone_reclaim() reclaims unmapped file backed pages and slab pages if over
defined limits as these are cheaper to reclaim.  The caller specifies the
order of the target allocation but the scan control is not being correctly
initialised.

The impact is that the correct number of pages are being reclaimed but
that lumpy reclaim is not being applied.  This increases the chances of a
full direct reclaim via try_to_free_pages() is required.

This patch initialises the order field of the scan control as requested by
the caller.

[mel@csn.ul.ie: rewrote changelog]
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: add comment why mark_page_accessed() would be better than pte_mkyoung() in follow...
KOSAKI Motohiro [Tue, 31 Mar 2009 22:19:37 +0000 (15:19 -0700)]
mm: add comment why mark_page_accessed() would be better than pte_mkyoung() in follow_page()

At first look, mark_page_accessed() in follow_page() seems a bit strange.
It seems pte_mkyoung() would be better consistent with other kernel code.

However, it is intentional. The commit log said:

    ------------------------------------------------
    commit 9e45f61d69be9024a2e6bef3831fb04d90fac7a8
    Author: akpm <akpm>
    Date:   Fri Aug 15 07:24:59 2003 +0000

    [PATCH] Use mark_page_accessed() in follow_page()

    Touching a page via follow_page() counts as a reference so we should be
    either setting the referenced bit in the pte or running mark_page_accessed().

    Altering the pte is tricky because we haven't implemented an atomic
    pte_mkyoung().  And mark_page_accessed() is better anyway because it has more
    aging state: it can move the page onto the active list.

    BKrev: 3f3c8acbplT8FbwBVGtth7QmnqWkIw
    ------------------------------------------------

The atomic issue is still true nowadays. adding comment help to understand
code intention and it would be better.

[akpm@linux-foundation.org: clarify text]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovmscan: clip swap_cluster_max in shrink_all_memory()
Johannes Weiner [Tue, 31 Mar 2009 22:19:35 +0000 (15:19 -0700)]
vmscan: clip swap_cluster_max in shrink_all_memory()

shrink_inactive_list() scans in sc->swap_cluster_max chunks until it hits
the scan limit it was passed.

shrink_inactive_list()
{
do {
isolate_pages(swap_cluster_max)
shrink_page_list()
} while (nr_scanned < max_scan);
}

This assumes that swap_cluster_max is not bigger than the scan limit
because the latter is checked only after at least one iteration.

In shrink_all_memory() sc->swap_cluster_max is initialized to the overall
reclaim goal in the beginning but not decreased while reclaim is making
progress which leads to subsequent calls to shrink_inactive_list()
reclaiming way too much in the one iteration that is done unconditionally.

Set sc->swap_cluster_max always to the proper goal before doing
  shrink_all_zones()
    shrink_list()
      shrink_inactive_list().

While the current shrink_all_memory() happily reclaims more than actually
requested, this patch fixes it to never exceed the goal:

unpatched
   wanted=10000 reclaimed=13356
   wanted=10000 reclaimed=19711
   wanted=10000 reclaimed=10289
   wanted=10000 reclaimed=17306
   wanted=10000 reclaimed=10700
   wanted=10000 reclaimed=10004
   wanted=10000 reclaimed=13301
   wanted=10000 reclaimed=10976
   wanted=10000 reclaimed=10605
   wanted=10000 reclaimed=10088
   wanted=10000 reclaimed=15000

patched
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=9599
   wanted=10000 reclaimed=8476
   wanted=10000 reclaimed=8326
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=9919
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=9624
   wanted=10000 reclaimed=10000
   wanted=10000 reclaimed=10000
   wanted=8500 reclaimed=8092
   wanted=316 reclaimed=316

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: MinChan Kim <minchan.kim@gmail.com>
Acked-by: Nigel Cunningham <ncunningham@crca.org.au>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
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>
15 years agomm: shrink_all_memory(): use sc.nr_reclaimed
MinChan Kim [Tue, 31 Mar 2009 22:19:34 +0000 (15:19 -0700)]
mm: shrink_all_memory(): use sc.nr_reclaimed

Commit a79311c14eae4bb946a97af25f3e1b17d625985d "vmscan: bail out of
direct reclaim after swap_cluster_max pages" moved the nr_reclaimed
counter into the scan control to accumulate the number of all reclaimed
pages in a reclaim invocation.

shrink_all_memory() can use the same mechanism. it increase code
consistency and redability.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: MinChan Kim <minchan.kim@gmail.com>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
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>
15 years agomm: don't call mark_page_accessed() in do_swap_page()
KOSAKI Motohiro [Tue, 31 Mar 2009 22:19:33 +0000 (15:19 -0700)]
mm: don't call mark_page_accessed() in do_swap_page()

commit bf3f3bc5e734706730c12a323f9b2068052aa1f0 (mm: don't
mark_page_accessed in fault path) only remove the mark_page_accessed() in
filemap_fault().

Therefore, swap-backed pages and file-backed pages have inconsistent
behavior.  mark_page_accessed() should be removed from do_swap_page().

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Hugh Dickins <hugh@veritas.com>
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>
15 years agomm: introduce for_each_populated_zone() macro
KOSAKI Motohiro [Tue, 31 Mar 2009 22:19:31 +0000 (15:19 -0700)]
mm: introduce for_each_populated_zone() macro

Impact: cleanup

In almost cases, for_each_zone() is used with populated_zone().  It's
because almost function doesn't need memoryless node information.
Therefore, for_each_populated_zone() can help to make code simplify.

This patch has no functional change.

[akpm@linux-foundation.org: small cleanup]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovmscan: rename sc.may_swap to may_unmap
Johannes Weiner [Tue, 31 Mar 2009 22:19:30 +0000 (15:19 -0700)]
vmscan: rename sc.may_swap to may_unmap

sc.may_swap does not only influence reclaiming of anon pages but pages
mapped into pagetables in general, which also includes mapped file pages.

In shrink_page_list():

if (!sc->may_swap && page_mapped(page))
goto keep_locked;

For anon pages, this makes sense as they are always mapped and reclaiming
them always requires swapping.

But mapped file pages are skipped here as well and it has nothing to do
with swapping.

The real effect of the knob is whether mapped pages are unmapped and
reclaimed or not.  Rename it to `may_unmap' to have its name match its
actual meaning more precisely.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: MinChan Kim <minchan.kim@gmail.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
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>
15 years agoget_mm_hiwater_xxx: trivial, s/define/inline/
Oleg Nesterov [Tue, 31 Mar 2009 22:19:29 +0000 (15:19 -0700)]
get_mm_hiwater_xxx: trivial, s/define/inline/

Andrew pointed out get_mm_hiwater_xxx() evaluate "mm" argument thrice/twice,
make them inline.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Hugh Dickins <hugh@veritas.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agooom_kill: don't call for int_sqrt(0)
Cyrill Gorcunov [Tue, 31 Mar 2009 22:19:27 +0000 (15:19 -0700)]
oom_kill: don't call for int_sqrt(0)

There is no need to call for int_sqrt if argument is 0.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Christoph Lameter <cl@linux-foundation.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovmap: remove needless lock and list in vmap
MinChan Kim [Tue, 31 Mar 2009 22:19:26 +0000 (15:19 -0700)]
vmap: remove needless lock and list in vmap

vmap's dirty_list is unused.  It's for optimizing flushing.  but Nick
didn't write the code yet.  so, we don't need it until time as it is
needed.

This patch removes vmap_block's dirty_list and codes related to it.

Signed-off-by: MinChan Kim <minchan.kim@gmail.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomm: mminit_validate_memmodel_limits(): remove redundant test
Cyrill Gorcunov [Tue, 31 Mar 2009 22:19:25 +0000 (15:19 -0700)]
mm: mminit_validate_memmodel_limits(): remove redundant test

In case if start_pfn overlap the upper bound no need to test end_pfn again
since we have it already trimmed.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoproc tty: remove struct tty_operations::read_proc
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:25 +0000 (15:19 -0700)]
proc tty: remove struct tty_operations::read_proc

struct tty_operations::proc_fops took it's place and there is one less
create_proc_read_entry() user now!

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch xtensa iss console to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:24 +0000 (15:19 -0700)]
proc tty: switch xtensa iss console to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch ia64 simserial to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:23 +0000 (15:19 -0700)]
proc tty: switch ia64 simserial to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch amiserial to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:23 +0000 (15:19 -0700)]
proc tty: switch amiserial to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch ircomm to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:22 +0000 (15:19 -0700)]
proc tty: switch ircomm to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch usb-serial to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:21 +0000 (15:19 -0700)]
proc tty: switch usb-serial to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch serial_core to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:21 +0000 (15:19 -0700)]
proc tty: switch serial_core to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch sdio_uart to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:20 +0000 (15:19 -0700)]
proc tty: switch sdio_uart to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch synclinkmp to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:20 +0000 (15:19 -0700)]
proc tty: switch synclinkmp to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch synclink_gt to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:19 +0000 (15:19 -0700)]
proc tty: switch synclink_gt to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch synclink to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:18 +0000 (15:19 -0700)]
proc tty: switch synclink to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch stallion to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:18 +0000 (15:19 -0700)]
proc tty: switch stallion to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch synclink_cs to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:17 +0000 (15:19 -0700)]
proc tty: switch synclink_cs to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch istallion to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:16 +0000 (15:19 -0700)]
proc tty: switch istallion to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch ip2 to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:16 +0000 (15:19 -0700)]
proc tty: switch ip2 to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoproc tty: switch cyclades to ->proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:15 +0000 (15:19 -0700)]
proc tty: switch cyclades to ->proc_fops

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: 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>
15 years agoproc tty: add struct tty_operations::proc_fops
Alexey Dobriyan [Tue, 31 Mar 2009 22:19:15 +0000 (15:19 -0700)]
proc tty: add struct tty_operations::proc_fops

Used for gradual switch of TTY drivers from using ->read_proc which helps
with gradual switch from ->read_proc for the whole tree.

As side effect, fix possible race condition when ->data initialized after
PDE is hooked into proc tree.

->proc_fops takes precedence over ->read_proc.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
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>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Tue, 31 Mar 2009 01:46:43 +0000 (18:46 -0700)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  wireless: remove duplicated .ndo_set_mac_address
  netfilter: xtables: fix IPv6 dependency in the cluster match
  tg3: Add GRO support.
  niu: Add GRO support.
  ucc_geth: Fix use-after-of_node_put() in ucc_geth_probe().
  gianfar: Fix use-after-of_node_put() in gfar_of_init().
  kernel: remove HIPQUAD()
  netpoll: store local and remote ip in net-endian
  netfilter: fix endian bug in conntrack printks
  dmascc: fix incomplete conversion to network_device_ops
  gso: Fix support for linear packets
  skbuff.h: fix missing kernel-doc
  ni5010: convert to net_device_ops

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Tue, 31 Mar 2009 01:46:12 +0000 (18:46 -0700)]
Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  sparc64: Fix reset hangs on Niagara systems.
  cpumask: use mm_cpumask() wrapper: sparc
  cpumask: remove dangerous CPU_MASK_ALL_PTR, &CPU_MASK_ALL.: sparc
  cpumask: remove the now-obsoleted pcibus_to_cpumask(): sparc
  cpumask: remove cpu_coregroup_map: sparc
  cpumask: prepare for iterators to only go to nr_cpu_ids/nr_cpumask_bits.: sparc
  cpumask: prepare for iterators to only go to nr_cpu_ids/nr_cpumask_bits.: sparc64
  cpumask: Use accessors code.: sparc64
  cpumask: Use accessors code: sparc
  cpumask: arch_send_call_function_ipi_mask: sparc
  cpumask: Use smp_call_function_many(): sparc64

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-cpumask
Linus Torvalds [Tue, 31 Mar 2009 01:00:26 +0000 (18:00 -0700)]
Merge git://git./linux/kernel/git/rusty/linux-2.6-cpumask

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-cpumask:
  oprofile: Thou shalt not call __exit functions from __init functions
  cpumask: remove the now-obsoleted pcibus_to_cpumask(): generic
  cpumask: remove cpumask_t from core
  cpumask: convert rcutorture.c
  cpumask: use new cpumask_ functions in core code.
  cpumask: remove references to struct irqaction's mask field.
  cpumask: use mm_cpumask() wrapper: kernel/fork.c
  cpumask: use set_cpu_active in init/main.c
  cpumask: remove node_to_first_cpu
  cpumask: fix seq_bitmap_*() functions.
  cpumask: remove dangerous CPU_MASK_ALL_PTR, &CPU_MASK_ALL

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-lguest-and-virtio
Linus Torvalds [Tue, 31 Mar 2009 00:57:39 +0000 (17:57 -0700)]
Merge git://git./linux/kernel/git/rusty/linux-2.6-lguest-and-virtio

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-lguest-and-virtio:
  lguest: barrier me harder
  lguest: use bool instead of int
  lguest: use KVM hypercalls
  lguest: wire up pte_update/pte_update_defer
  lguest: fix spurious BUG_ON() on invalid guest stack.
  virtio: more neatening of virtio_ring macros.
  virtio: fix BAD_RING, START_US and END_USE macros

15 years agoMerge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6
Linus Torvalds [Tue, 31 Mar 2009 00:54:32 +0000 (17:54 -0700)]
Merge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6

* 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6:
  hwmon: (fschmd) Add support for the FSC Hades IC
  hwmon: (fschmd) Add support for the FSC Syleus IC
  i2c-i801: Instantiate FSC hardware montioring chips
  dmi: Let dmi_walk() users pass private data
  hwmon: Define a standard interface for chassis intrusion detection
  Move the pcf8591 driver to hwmon
  hwmon: (w83627ehf) Only expose in6 or temp3 on the W83667HG
  hwmon: (w83627ehf) Add support for W83667HG
  hwmon: (w83627ehf) Invert fan pin variables logic
  hwmon: (hdaps) Fix Thinkpad X41 axis inversion
  hwmon: (hdaps) Allow inversion of separate axis
  hwmon: (ds1621) Clean up documentation
  hwmon: (ds1621) Avoid unneeded register access
  hwmon: (ds1621) Clean up register access
  hwmon: (ds1621) Reorder code statements

15 years agoMerge branch 'locking-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 31 Mar 2009 00:17:35 +0000 (17:17 -0700)]
Merge branch 'locking-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'locking-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (33 commits)
  lockdep: fix deadlock in lockdep_trace_alloc
  lockdep: annotate reclaim context (__GFP_NOFS), fix SLOB
  lockdep: annotate reclaim context (__GFP_NOFS), fix
  lockdep: build fix for !PROVE_LOCKING
  lockstat: warn about disabled lock debugging
  lockdep: use stringify.h
  lockdep: simplify check_prev_add_irq()
  lockdep: get_user_chars() redo
  lockdep: simplify get_user_chars()
  lockdep: add comments to mark_lock_irq()
  lockdep: remove macro usage from mark_held_locks()
  lockdep: fully reduce mark_lock_irq()
  lockdep: merge the !_READ mark_lock_irq() helpers
  lockdep: merge the _READ mark_lock_irq() helpers
  lockdep: simplify mark_lock_irq() helpers #3
  lockdep: further simplify mark_lock_irq() helpers
  lockdep: simplify the mark_lock_irq() helpers
  lockdep: split up mark_lock_irq()
  lockdep: generate usage strings
  lockdep: generate the state bit definitions
  ...

15 years agoMerge branch 'proc-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/adobriyan...
Linus Torvalds [Mon, 30 Mar 2009 23:06:04 +0000 (16:06 -0700)]
Merge branch 'proc-linus' of git://git./linux/kernel/git/adobriyan/proc

* 'proc-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/adobriyan/proc:
  Revert "proc: revert /proc/uptime to ->read_proc hook"
  proc 2/2: remove struct proc_dir_entry::owner
  proc 1/2: do PDE usecounting even for ->read_proc, ->write_proc
  proc: fix sparse warnings in pagemap_read()
  proc: move fs/proc/inode-alloc.txt comment into a source file

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Mon, 30 Mar 2009 22:12:14 +0000 (15:12 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/rafael/suspend-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/suspend-2.6:
  PCI PM: Make pci_prepare_to_sleep() disable wake-up if needed
  radeonfb: Use __pci_complete_power_transition()
  PCI PM: Introduce __pci_[start|complete]_power_transition() (rev. 2)
  PCI PM: Restore config spaces of all devices during early resume
  PCI PM: Make pci_set_power_state() handle devices with no PM support
  PCI PM: Put devices into low power states during late suspend (rev. 2)
  PCI PM: Move pci_restore_standard_config to pci-driver.c
  PCI PM: Use pci_set_power_state during early resume
  PCI PM: Consistently use variable name "error" for pm call return values
  kexec: Change kexec jump code ordering
  PM: Change hibernation code ordering
  PM: Change suspend code ordering
  PM: Rework handling of interrupts during suspend-resume
  PM: Introduce functions for suspending and resuming device interrupts

15 years agodma-debug: fix printk formats (i386)
Randy Dunlap [Mon, 30 Mar 2009 21:08:44 +0000 (14:08 -0700)]
dma-debug: fix printk formats (i386)

Fix printk format warnings in dma-debug:

  lib/dma-debug.c:645: warning: format '%016llx' expects type 'long long unsigned int', but argument 6 has type 'dma_addr_t'
  lib/dma-debug.c:662: warning: format '%016llx' expects type 'long long unsigned int', but argument 6 has type 'dma_addr_t'
  lib/dma-debug.c:676: warning: format '%016llx' expects type 'long long unsigned int', but argument 6 has type 'dma_addr_t'
  lib/dma-debug.c:686: warning: format '%016llx' expects type 'long long unsigned int', but argument 6 has type 'dma_addr_t'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoreiserfs: xattr_create is unused with xattrs disabled
Jeff Mahoney [Mon, 30 Mar 2009 20:49:58 +0000 (16:49 -0400)]
reiserfs: xattr_create is unused with xattrs disabled

This patch ifdefs xattr_create when xattrs aren't enabled.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoreiserfs: fix build breakage
Alexander Beregalov [Mon, 30 Mar 2009 20:32:40 +0000 (00:32 +0400)]
reiserfs: fix build breakage

Fix this build error when REISERFS_FS_POSIX_ACL is not set:

  fs/reiserfs/inode.c: In function 'reiserfs_new_inode':
  fs/reiserfs/inode.c:1919: warning: passing argument 1 of 'reiserfs_inherit_default_acl' from incompatible pointer type
  fs/reiserfs/inode.c:1919: warning: passing argument 2 of 'reiserfs_inherit_default_acl' from incompatible pointer type
  fs/reiserfs/inode.c:1919: warning: passing argument 3 of 'reiserfs_inherit_default_acl' from incompatible pointer type
  fs/reiserfs/inode.c:1919: error: too many arguments to function 'reiserfs_inherit_default_acl'

due to a missing transaction-handle argument in the non-acl
compatibility function.

Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Acked-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agolockdep: fix deadlock in lockdep_trace_alloc
Peter Zijlstra [Fri, 20 Mar 2009 10:13:20 +0000 (11:13 +0100)]
lockdep: fix deadlock in lockdep_trace_alloc

Heiko reported that we grab the graph lock with irqs enabled.

Fix this by providng the same wrapper as all other lockdep entry
functions have.

Reported-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Nick Piggin <npiggin@suse.de>
LKML-Reference: <1237544000.24626.52.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoRevert "proc: revert /proc/uptime to ->read_proc hook"
Alexey Dobriyan [Fri, 20 Feb 2009 14:07:22 +0000 (17:07 +0300)]
Revert "proc: revert /proc/uptime to ->read_proc hook"

This reverts commit 6c87df37dcb9c6c33923707fa5191e0a65874d60.

proc files implemented through seq_file do pread(2) now.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
15 years agoproc 2/2: remove struct proc_dir_entry::owner
Alexey Dobriyan [Wed, 25 Mar 2009 19:48:06 +0000 (22:48 +0300)]
proc 2/2: remove struct proc_dir_entry::owner

Setting ->owner as done currently (pde->owner = THIS_MODULE) is racy
as correctly noted at bug #12454. Someone can lookup entry with NULL
->owner, thus not pinning enything, and release it later resulting
in module refcount underflow.

We can keep ->owner and supply it at registration time like ->proc_fops
and ->data.

But this leaves ->owner as easy-manipulative field (just one C assignment)
and somebody will forget to unpin previous/pin current module when
switching ->owner. ->proc_fops is declared as "const" which should give
some thoughts.

->read_proc/->write_proc were just fixed to not require ->owner for
protection.

rmmod'ed directories will be empty and return "." and ".." -- no harm.
And directories with tricky enough readdir and lookup shouldn't be modular.
We definitely don't want such modular code.

Removing ->owner will also make PDE smaller.

So, let's nuke it.

Kudos to Jeff Layton for reminding about this, let's say, oversight.

http://bugzilla.kernel.org/show_bug.cgi?id=12454

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
15 years agoproc 1/2: do PDE usecounting even for ->read_proc, ->write_proc
Alexey Dobriyan [Fri, 20 Feb 2009 14:04:33 +0000 (17:04 +0300)]
proc 1/2: do PDE usecounting even for ->read_proc, ->write_proc

struct proc_dir_entry::owner is going to be removed. Now it's only necessary
to protect PDEs which are using ->read_proc, ->write_proc hooks.

However, ->owner assignments are racy and make it very easy for someone to switch
->owner on live PDE (as some subsystems do) without fixing refcounts and so on.

http://bugzilla.kernel.org/show_bug.cgi?id=12454

So, ->owner is on death row.

Proxy file operations exist already (proc_file_operations), just bump usecount
when necessary.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
15 years agoproc: fix sparse warnings in pagemap_read()
Milind Arun Choudhary [Fri, 20 Feb 2009 13:56:45 +0000 (16:56 +0300)]
proc: fix sparse warnings in pagemap_read()

fs/proc/task_mmu.c:696:12: warning: cast removes address space of expression
fs/proc/task_mmu.c:696:9: warning: incorrect type in assignment (different address spaces)
fs/proc/task_mmu.c:696:9:    expected unsigned long long [noderef] [usertype] <asn:1>*out
fs/proc/task_mmu.c:696:9:    got unsigned long long [usertype] *<noident>
fs/proc/task_mmu.c:697:12: warning: cast removes address space of expression
fs/proc/task_mmu.c:697:9: warning: incorrect type in assignment (different address spaces)
fs/proc/task_mmu.c:697:9:    expected unsigned long long [noderef] [usertype] <asn:1>*end
fs/proc/task_mmu.c:697:9:    got unsigned long long [usertype] *<noident>
fs/proc/task_mmu.c:723:12: warning: cast removes address space of expression
fs/proc/task_mmu.c:723:26: error: subtraction of different types can't work (different address spaces)
fs/proc/task_mmu.c:725:24: error: subtraction of different types can't work (different address spaces)

Signed-off-by: Milind Arun Choudhary <milindchoudhary@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
15 years agoproc: move fs/proc/inode-alloc.txt comment into a source file
Randy Dunlap [Tue, 13 Jan 2009 10:53:48 +0000 (13:53 +0300)]
proc: move fs/proc/inode-alloc.txt comment into a source file

so that people will realize that it exists and can update it as needed.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>