Andrew Morton [Fri, 10 Aug 2007 20:01:06 +0000 (13:01 -0700)]
mtdchar build fix
sh:
drivers/mtd/mtdchar.c: In function `mtd_mmap':
drivers/mtd/mtdchar.c:817: error: dereferencing pointer to incomplete type
drivers/mtd/mtdchar.c:817: error: `VM_SHARED' undeclared (first use in this function)
drivers/mtd/mtdchar.c:817: error: (Each undeclared identifier is reported only once
Cc: David Woodhouse <dwmw2@infradead.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The problem can be triggered when tcflush() is called when data are being
pushed to the line discipline driver by flush_to_ldisc().
flush_to_ldisc() releases tty->buf.lock when calling the line discipline
receive_buf function. At that poing tty_buffer_flush() kicks in and sets both
tty->buf.head and tty->buf.tail to NULL. When flush_to_ldisc() finishes, it
restores tty->buf.head but doesn't touch tty->buf.tail. This corrups the
buffer queue, and the next call to tty_buffer_request_room() will allocate a
new buffer and overwrite tty->buf.head. The previous buffer is then lost
forever without being released.
(Thanks to Laurent for the above text, for finding, disgnosing and reporting
the bug)
- Use tty->flags bits for the flush status.
- Wait for the flag to clear again before returning
- Fix the doc error noted
- Fix flush of empty queue leaving stale flushpending
[akpm@linux-foundation.org: cleanup] Signed-off-by: Alan Cox <alan@redhat.com> Acked-by: Paul Fulghum <paulkf@microgate.com> Cc: Laurent Pinchart <laurentp@cse-semaphore.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Jesper Juhl [Fri, 10 Aug 2007 20:01:04 +0000 (13:01 -0700)]
Documentation: sysrq, description of 'h' slightly inaccurate
In Documentation/sysrq.txt, the description of 'h' says that any key not
listed *above* will generate help. That's obviously not true since all the
keys listed below 'h' will do what they are described to do, not display help.
So change the text so that it says that any key not listed in the table will
generate help, which is what really happens.
Andy Whitcroft [Fri, 10 Aug 2007 20:01:03 +0000 (13:01 -0700)]
update checkpatch.pl to version 0.09
This version brings a number of new checks, and a number of bug
fixes. Of note:
- checks for spacing on round and square bracket combinations
- loosening of the single statement brace checks, to allow
them when they contain comments or where other blocks in a
compound statement have them.
- parks the multple declaration support
- allows architecture defines in architecture specific headers
Andy Whitcroft (21):
Version: 0.09
loosen single statement brace checks
fix up multiple declaration to avoid function arguments
add some function space parenthesis check exceptions
handle EXPORT_'s with parentheses in their names
clean up some warnings in multi-line macro bracketing support
park the multiple declaration checks
make block brace checks count comments as a statement
__volatile__ and __extension__ are not functions
allow architecture specific defined within architecture includes
check spacing on square brackets
check spacing on parentheses
ensure we apply checks to the part before start comment
check #ifdef conditional spacing
handle __init_refok and __must_check
add noinline to inline checks
prevent email addresses from tripping spacing checks
handle typed initialiser spacing
handle line contination as end of line
add bool to the type matcher
refine EXPORT_SYMBOL checks to handle pointers
Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
While specs says: "The SPI baud rate generator clock source (either
system clock or system clock divided by 16, depending on DIV16 bit) is
divided by 4 * ([PM] + 1), a range from 4 to 64.".
Thus " - 1" is missing in the spi_mpc83xx's formula.
Why nobody noticed that bug? Probably because sysclk usually less then
user expects, e.g. you expect 200 MHz, but real clock is 198 MHz,
and integer rounding helps when this formula is used.
Suppose it's SPI in QE, SYSCLK at 198 MHz, thus SPIBRG at 99MHz, 25 MHz
requested.
PM = (99MHz / ( 25 MHz * 4 )), PM == 0, output SPICLK will be 24.75 MHz
At lower frequencies this bug is more noticeable, though.
And this bug shows itself in all its beauty if SYSCLK is equal or a bit
more than you expect (200 MHz SYSCLK, 100 MHz SPIBRG):
PM = (100MHz / ( 25 MHz * 4 )), PM == 1, output SPICLK will be 12.625 MHz!
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com> 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>
Anton Vorontsov [Fri, 10 Aug 2007 20:01:01 +0000 (13:01 -0700)]
spi_mpc83xx: in "QE mode", use sysclk/2
For MPC8349E input to the SPI Baud Rate Generator is SYSCLK, but it's
SYSCLK/2 for MPC8323E (SPI in QE). Fix this, and remove confusion by
renaming the mpc83xx_spi->sysclk member as mpc83xx_spi->spibrg.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com> 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>
Hibernation: do not try to mark invalid PFNs as nosave
On some systems some PFNs reported by the early initialization code as
'nosave' may be invalid. If we try to set the corresponding bits in the
hibernation bitmap, BUG_ON() in memory_bm_find_bit() will be triggered and
the system won't be able to boot (cf.
https://bugzilla.novell.com/show_bug.cgi?id=296242).
Prevent this from happening by verifying if the 'nosave' PFNs are valid in
mark_nosave_pages().
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ryusuke Konishi [Fri, 10 Aug 2007 20:00:56 +0000 (13:00 -0700)]
eCryptfs: fix error handling in ecryptfs_init
ecryptfs_init() exits without doing any cleanup jobs if
ecryptfs_init_messaging() fails. In that case, eCryptfs leaves
sysfs entries, leaks memory, and causes an invalid page fault.
This patch fixes the problem.
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Michael Halcrow <mhalcrow@us.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Gabriel C [Fri, 10 Aug 2007 20:00:56 +0000 (13:00 -0700)]
linux-audit list is subscribers-only
Signed-off-by: Gabriel Craciunescu <nix.or.die@googlemail.com> Cc: David Woodhouse <dwmw2@infradead.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ryusuke Konishi [Fri, 10 Aug 2007 20:00:51 +0000 (13:00 -0700)]
eCryptfs: fix lookup error for special files
When ecryptfs_lookup() is called against special files, eCryptfs generates
the following errors because it tries to treat them like regular eCryptfs
files.
Error opening lower file for lower_dentry [0xffff810233a6f150], lower_mnt [0xffff810235bb4c80], and flags
[0x8000]
Error opening lower_file to read header region
Error attempting to read the [user.ecryptfs] xattr from the lower file; return value = [-95]
Valid metadata not found in header region or xattr region; treating file as unencrypted
For instance, the problem can be reproduced by the steps below.
# mkdir /root/crypt /mnt/crypt
# mount -t ecryptfs /root/crypt /mnt/crypt
# mknod /mnt/crypt/c0 c 0 0
# umount /mnt/crypt
# mount -t ecryptfs /root/crypt /mnt/crypt
# ls -l /mnt/crypt
This patch fixes it by adding a check similar to directories and
symlinks.
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Acked-by: Michael Halcrow <mhalcrow@us.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Paul A. Clarke [Fri, 10 Aug 2007 20:00:49 +0000 (13:00 -0700)]
matroxfb: rectify jitter (G450/G550)
This builds upon my previous attempts to resolve some jitter problems seen
with the Matrox G450 and G550 -based cards, including odd disparities observed
between x86 and Power -based machines in a somewhat less hackish way (removing
the hacked ifdefs).
Apparently, preference should be given to use the DVI PLL when frequencies
permit, the Standard PLL otherwise. The max pixel clock for the panellink
interface is extracted from the PInS information on the card and used as a
limit to determine which PLL to use.
Signed-off-by: Paul A. Clarke <pc@us.ibm.com> Acked-by: Petr Vandrovec <petr@vandrovec.name> 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>
Adrian McMenamin [Fri, 10 Aug 2007 20:00:48 +0000 (13:00 -0700)]
pvr2fb: Consolidated cleanup of pvr2fb.c
- better handling of the pvr2 registers based on more up to date information.
Testing shows that it seems to work pretty well at 16bpp, 24bpp and 32bpp -
including proper rendering of the boot logo at all levels (previously this was
a bit broken even at 16bpp) and giving white against black text. Really
detailed testing (eg with X11) requires support for the maple bus - which
isn't (currently - next project assuming this is okay) available, but I have
no reason to think this is broken.
Signed-off by: Adrian McMenamin <adrian@mcmen.demon.co.uk> Acked-by: Paul Mundt <lethal@linux-sh.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>
Reported by: Adrian McMenamin <adrianmcmenamin@gmail.com>
This driver will oops when the pseudo_palette[] is written as u32 but not when
written as u16. When written as u32, it corrupts the adjacent 'mmio_base'
field of struct pvr2fb_par. Fix by using framebuffer_alloc()/release() to
allocate struct fb_info and struct pvr2fb_par, and create the pseudo_palette[]
as part of struct pvr2fb_par.
Helge Deller [Fri, 10 Aug 2007 20:00:45 +0000 (13:00 -0700)]
stifb: detect cards in double buffer mode more reliably
Visualize-EG, Graffiti and A4450A graphics cards on PARISC can
be configured in double-buffer and standard mode, but the stifb
driver supports standard mode only.
This patch detects double-buffered cards more reliable.
It is a real bugfix for a very nasty problem for all parisc users which have
wrongly configured their graphic card. The problem: The stifb graphics driver
will not detect that the card is wrongly configured and then nevertheless just
enables the graphics mode, which it shouldn't. In the end, the user will see
no further updates / boot messages on the screen.
We had documented this problem already on our FAQ
(http://parisc-linux.org/faq/index.html#viseg "Why do I get corrupted graphics
with my Vis-EG/Graffiti/A4450A card?") but people still run into this problem.
So having this fix in as early as possible can help us.
Badari Pulavarty [Fri, 10 Aug 2007 20:00:44 +0000 (13:00 -0700)]
direct-io: fix error-path crashes
Need to initialize map_bh.b_state to zero. Otherwise, in case of a faulty
user-buffer its possible to go into dio_zero_block() and submit a page by
mistake - since it checks for buffer_new().
Robin Holt [Fri, 10 Aug 2007 20:00:43 +0000 (13:00 -0700)]
x86_64: fix HPET init race
I have had four seperate system lockups attributable to this exact problem
in two days of testing. Instead of trying to handle all the weird end
cases and wrap, how about changing it to look for exactly what we appear
to want.
The following patch removes a couple races in setup_APIC_timer. One occurs
when the HPET advances the COUNTER past the T0_CMP value between the time
the T0_CMP was originally read and when COUNTER is read. This results in
a delay waiting for the counter to wrap. The other results from the counter
wrapping.
This change takes a snapshot of T0_CMP at the beginning of the loop and
simply loops until T0_CMP has changed (a tick has happened).
<later>
I have one small concern about the patch. I am not sure it meets the intent
as well as it should. I think we are trying to match APIC timer interrupts up
with the hpet counter increment. The event which appears to be disturbing
this loop in our test environment is the NMI watchdog. What we believe has
been happening with the existing code is the setup_APIC_timer loop has read
the CMP value, and the NMI watchdog code fires for the first time. This
results in a series of icache miss slowdowns and by the time we get back to
things it has wrapped.
I think this code is trying to get the CMP as close to the counter value as
possible. If that is the intent, maybe we should really be testing against a
"window" around the CMP. Something like COUNTER = CMP+/2. It appears COUNTER
should get advanced every 89nSec (IIRC). The above seems like an unreasonably
small window, but may be necessary. Without documentation, I am not sure of
the original intent with this code.
In summary, this code fixes my boot hangs, but since I am not certain of the
intent of the existing code, I am not certain this has not introduced new bugs
or unexpected behaviors.
Jesper Juhl [Wed, 8 Aug 2007 23:31:30 +0000 (16:31 -0700)]
SLUB: Fix format specifier in Documentation/vm/slabinfo.c
There's a little problem in Documentation/vm/slabinfo.c
The code is using "%d" in a printf() call to print an 'unsigned long'.
This patch corrects it to use "%lu" instead.
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com> Signed-off-by: Christoph Lameter <clameter@sgi.com>
The dynamic dma kmalloc creation can run into trouble if a
GFP_ATOMIC allocation is the first one performed for a certain size
of dma kmalloc slab.
- Move the adding of the slab to sysfs into a workqueue
(sysfs does GFP_KERNEL allocations)
- Do not call kmem_cache_destroy() (uses slub_lock)
- Only acquire the slub_lock once and--if we cannot wait--do a trylock.
This introduces a slight risk of the first kmalloc(x, GFP_DMA|GFP_ATOMIC)
for a range of sizes failing due to another process holding the slub_lock.
However, we only need to acquire the spinlock once in order to establish
each power of two DMA kmalloc cache. The possible conflict is with the
slub_lock taken during slab management actions (create / remove slab cache).
It is rather typical that a driver will first fill its buffers using
GFP_KERNEL allocations which will wait until the slub_lock can be acquired.
Drivers will also create its slab caches first outside of an atomic
context before starting to use atomic kmalloc from an interrupt context.
If there are any failures then they will occur early after boot or when
loading of multiple drivers concurrently. Drivers can already accomodate
failures of GFP_ATOMIC for other reasons. Retries will then create the slab.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
SLUB: Remove checks for MAX_PARTIAL from kmem_cache_shrink
The MAX_PARTIAL checks were supposed to be an optimization. However, slab
shrinking is a manually triggered process either through running slabinfo
or by the kernel calling kmem_cache_shrink.
If one really wants to shrink a slab then all operations should be done
regardless of the size of the partial list. This also fixes an issue that
could surface if the number of partial slabs was initially above MAX_PARTIAL
in kmem_cache_shrink and later drops below MAX_PARTIAL through the
elimination of empty slabs on the partial list (rare). In that case a few
slabs may be left off the partial list (and only be put back when they
are empty).
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Alan Cox [Wed, 8 Aug 2007 23:57:54 +0000 (00:57 +0100)]
remove dubious legal statment from uio-howto
UIO currently contains a rather dubious statement which wants removing.
The actual questions around whether user space code that depends tightly
on kernel GPL code designed to co-work with it are derivative works of
the kernel is extremely complex, and since we don't have space for either
a masters length essay on legal issues or need to start flamewars lets
simply remove the comment and leave law to lawyers
Signed-off-by: Alan Cox <alan@redhat.com> Signed-off-by: Linus Torvalds <tovalds@linux-foundation.org>
Linus Torvalds [Thu, 9 Aug 2007 15:38:14 +0000 (08:38 -0700)]
Merge git://git.linux-nfs.org/pub/linux/nfs-2.6
* git://git.linux-nfs.org/pub/linux/nfs-2.6:
SUNRPC: Replace flush_workqueue() with cancel_work_sync() and friends
NFS: Replace flush_scheduled_work with cancel_work_sync() and friends
SUNRPC: Don't call gss_delete_sec_context() from an rcu context
NFSv4: Don't call put_rpccred() from an rcu callback
NFS: Fix NFSv4 open stateid regressions
NFSv4: Fix a locking regression in nfs4_set_mode_locked()
NFS: Fix put_nfs_open_context
SUNRPC: Fix a race in rpciod_down()
Linus Torvalds [Thu, 9 Aug 2007 15:27:25 +0000 (08:27 -0700)]
Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6
* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
[SPARC64]: Fix memory leak when cpu hotplugging.
[SPARC64]: Do not assume sun4v chips have load-twin/store-init support.
[SPARC64]: Fix hard-coding of cpu type output in /proc/cpuinfo on sun4v.
[SPARC]: Centralize find_in_proplist() instead of duplicating N times.
Linus Torvalds [Thu, 9 Aug 2007 15:23:47 +0000 (08:23 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
mmc: at91_mci: remove whitespace at the end of lines
mmc: reorganize bounce buffer init
wbsd: fix section mismatch warnings
* git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched: (61 commits)
sched: refine negative nice level granularity
sched: fix update_stats_enqueue() reniced codepath
sched: round a bit better
sched: make the multiplication table more accurate
sched: optimize update_rq_clock() calls in the load-balancer
sched: optimize activate_task()
sched: clean up set_curr_task_fair()
sched: remove __update_rq_clock() call from entity_tick()
sched: move the __update_rq_clock() call to scheduler_tick()
sched debug: remove the 'u64 now' parameter from print_task()/_rq()
sched: remove the 'u64 now' local variables
sched: remove the 'u64 now' parameter from deactivate_task()
sched: remove the 'u64 now' parameter from dequeue_task()
sched: remove the 'u64 now' parameter from enqueue_task()
sched: remove the 'u64 now' parameter from dec_nr_running()
sched: remove the 'u64 now' parameter from inc_nr_running()
sched: remove the 'u64 now' parameter from dec_load()
sched: remove the 'u64 now' parameter from inc_load()
sched: remove the 'u64 now' parameter from update_curr_load()
sched: remove the 'u64 now' parameter from ->task_new()
...
lguest: avoid shared libraries mapped over guest memory
Some versions of ld.so mmap the shared libraries right in over guest
memory, so compile lguest statically by default.
[ FC7 maps shared libraries very low, where the launcher maps guest's
physical memory. Quick fix is to link Launcher static, real fix is
for 2.6.24. ]
-static is a simple fix. I expect this problem will be more common than we
like, as different distro's make different "improvements" to ld.so
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Rusty Russell [Thu, 9 Aug 2007 10:57:13 +0000 (20:57 +1000)]
lguest: Fix Malicious Guest GDT Host Crash
If a Guest makes hypercall which sets a GDT entry to not present, we
currently set any segment registers using that GDT entry to 0.
Unfortunately, this is not sufficient: there are other ways of
altering GDT entries which will cause a fault.
The correct solution to do what Linux does: let them set any GDT value
they want and handle the #GP when popping causes a fault. This has
the added benefit of making our Switcher slightly more robust in the
case of any other bugs which cause it to fault.
We kill the Guest if it causes a fault in the Switcher: it's the
Guest's responsibility to make sure it's not using segments when it
changes them.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Rusty Russell [Thu, 9 Aug 2007 10:52:35 +0000 (20:52 +1000)]
Fix non-TSC guest clocksource lockup
lguest uses a host-supplied wallclock-based clocksource when the TSC
is not reliable. As this is already in nanoseconds, I naively used a
multiplier of 1 and a shift of 0.
But update_wall_time() in its infinite wisdom decides to adjust the
clock a little (where does it think it's getting a more accurate time
from?)
It will happily tweak the multiplier... to 0, then -1.
So the "fix" is to use a shift of 22 like everyone else, and a
multiplier of 1 << 22.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Thu, 9 Aug 2007 15:10:16 +0000 (08:10 -0700)]
Revert "genirq: temporary fix for level-triggered IRQ resend"
This reverts commit 0fc4969b866671dfe39b1a9119d0fdc7ea0f63e5. It was
always meant to be temporary, but it's generating more useless noise
than anything else, and we probably should never have done it in the
generic kernel (only had the people involved test it on their own).
Gabriel C [Thu, 2 Aug 2007 18:20:44 +0000 (20:20 +0200)]
wbsd: fix section mismatch warnings
This patch fixes the following section mismatch warnings
...
WARNING: vmlinux.o(.init.text+0x29d40): Section mismatch: reference to .exit.text:wbsd_release_resources (between 'wbsd_init' and 'wbsd_probe')
WARNING: vmlinux.o(.init.text+0x29d49): Section mismatch: reference to .exit.text:wbsd_free_mmc (between 'wbsd_init' and 'wbsd_probe')
WARNING: vmlinux.o(.init.text+0x29f28): Section mismatch: reference to .exit.text:wbsd_free_mmc (between 'wbsd_init' and 'wbsd_probe')
...
Signed-off-by: Gabriel Craciunescu <nix.or.die@googlemail.com> Acked-by: Sam Ravnborg <sam@ravnborg.org> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Ingo Molnar [Thu, 9 Aug 2007 09:16:52 +0000 (11:16 +0200)]
sched: refine negative nice level granularity
refine the granularity of negative nice level tasks: let them
reschedule more often to offset the effect of them consuming
their wait_runtime proportionately slower. (This makes nice-0
task scheduling smoother in the presence of negatively
reniced tasks.)
Ingo Molnar [Thu, 9 Aug 2007 09:16:51 +0000 (11:16 +0200)]
sched: optimize update_rq_clock() calls in the load-balancer
optimize update_rq_clock() calls in the load-balancer: update them
right after locking the runqueue(s) so that the pull functions do
not have to call it.
Ingo Molnar [Thu, 9 Aug 2007 09:16:51 +0000 (11:16 +0200)]
sched: optimize activate_task()
optimize activate_task() by removing update_rq_clock() from it.
(and add update_rq_clock() to all callsites of activate_task() that
did not have it before.)
Ingo Molnar [Thu, 9 Aug 2007 09:16:47 +0000 (11:16 +0200)]
sched: remove 'now' use from assignments
change all 'now' timestamp uses in assignments to rq->clock.
( this is an identity transformation that causes no functionality change:
all such new rq->clock is necessarily preceded by an update_rq_clock()
call. )
Ingo Molnar [Thu, 9 Aug 2007 09:16:46 +0000 (11:16 +0200)]
sched: add [__]update_rq_clock(rq)
add the [__]update_rq_clock(rq) functions. (No change in functionality,
just reorganization to prepare for elimination of the heavy 64-bit
timestamp-passing in the scheduler.)
Peter Williams [Thu, 9 Aug 2007 09:16:46 +0000 (11:16 +0200)]
sched: fix bug in balance_tasks()
There are two problems with balance_tasks() and how it used:
1. The variables best_prio and best_prio_seen (inherited from the old
move_tasks()) were only required to handle problems caused by the
active/expired arrays, the order in which they were processed and the
possibility that the task with the highest priority could be on either.
These issues are no longer present and the extra overhead associated
with their use is unnecessary (and possibly wrong).
2. In the absence of CONFIG_FAIR_GROUP_SCHED being set, the same
this_best_prio variable needs to be used by all scheduling classes or
there is a risk of moving too much load. E.g. if the highest priority
task on this at the beginning is a fairly low priority task and the rt
class migrates a task (during its turn) then that moved task becomes the
new highest priority task on this_rq but when the sched_fair class
initializes its copy of this_best_prio it will get the priority of the
original highest priority task as, due to the run queue locks being
held, the reschedule triggered by pull_task() will not have taken place.
This could result in inappropriate overriding of skip_for_load and
excessive load being moved.
The attached patch addresses these problems by deleting all reference to
best_prio and best_prio_seen and making this_best_prio a reference
parameter to the various functions involved.
load_balance_fair() has also been modified so that this_best_prio is
only reset (in the loop) if CONFIG_FAIR_GROUP_SCHED is set. This should
preserve the effect of helping spread groups' higher priority tasks
around the available CPUs while improving system performance when
CONFIG_FAIR_GROUP_SCHED isn't set.
Signed-off-by: Peter Williams <pwil3058@bigpond.net.au> Signed-off-by: Ingo Molnar <mingo@elte.hu>
projects / linux-2.6 / log