]> pilppa.com Git - linux-2.6-omap-h63xx.git/log
linux-2.6-omap-h63xx.git
17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davej/agpgart
Linus Torvalds [Thu, 21 Jun 2007 22:59:11 +0000 (15:59 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/davej/agpgart

* master.kernel.org:/pub/scm/linux/kernel/git/davej/agpgart:
  [AGPGART] intel_agp: don't load if no IGD and AGP port

17 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kyle/parisc-2.6
Linus Torvalds [Thu, 21 Jun 2007 22:57:50 +0000 (15:57 -0700)]
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kyle/parisc-2.6

* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kyle/parisc-2.6:
  [PARISC] unwinder improvements
  [PARISC] Fix unwinder on 64-bit kernels
  [PARISC] Handle wrapping in expand_upwards()
  [PARISC] stop lcd driver from stripping initial whitespace

17 years agoposix-timers: Prevent softirq starvation by small intervals and SIG_IGN
Thomas Gleixner [Thu, 21 Jun 2007 20:45:15 +0000 (20:45 +0000)]
posix-timers: Prevent softirq starvation by small intervals and SIG_IGN

posix-timers which deliver an ignored signal are currently rearmed in
the timer softirq: This is necessary because the timer needs to be
delivered again when SIG_IGN is removed. This is not a problem, when
the interval is reasonable.

With high resolution timers enabled one might arm a posix timer with a
very small interval and ignore the signal. This might lead to a
softirq starvation when the interval is so small that the timer is
requeued onto the softirq pending list right away.

This problem was pointed out by Jan Kiszka. Thanks Jan !

The correct solution would be to stop the timer, when the signal is
ignored and rearm it when SIG_IGN is removed. Unfortunately this
requires modification in sigaction and involves non trivial sighand
locking. It's too late in the release cycle for such a change.

For now we just keep the timer running and enforce that the timer only
fires every jiffie. This does not break anything as we keep the
overrun counter correct. It adds a little inaccuracy to the
timer_gettime() interface, but...

The more complex change is necessary anyway to fix another short
coming of the current implementation, which I discovered while looking
at this problem: A pending signal is discarded when SIG_IGN is set. In
case that a posixtimer signal is pending then it is discarded as well,
but when SIG_IGN is removed later nothing rearms the timer. This is
not new, it's that way since posix timers have been merged. So nothing
to worry about right now.

I have a working solution to fix all of this, but the impact is too
large for both stable and 2.6.22. I'm going to send it out for review
in the next days.

This should go into 2.6.21.stable as well.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Jan Kiszka <jan.kiszka@web.de>
Cc: Ulrich Drepper <drepper@redhat.com>
Cc: Stable Team <stable@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agopage_mapping must avoid slub pages
Hugh Dickins [Thu, 21 Jun 2007 22:27:45 +0000 (23:27 +0100)]
page_mapping must avoid slub pages

Nicolas Ferre reports oops from flush_dcache_page() on ARM when using
SLUB: which reuses page->mapping as page->slab.  The page_mapping()
function, used by ARM and PA-RISC flush_dcache_page() implementations,
must not confuse SLUB pages with those which have page->mapping set.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Nicolas Ferre <nicolas.ferre@rfo.atmel.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PARISC] unwinder improvements
Randolph Chung [Tue, 12 Jun 2007 06:27:32 +0000 (14:27 +0800)]
[PARISC] unwinder improvements

Add special-case handling for "handle_interruption" so that we can rewind
past the interruption. This is useful for seeing what caused a BUG() or
WARN_ON(); otherwise the unwind stops at the interruption.

Signed-off-by: Randolph Chung <tausq@debian.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
17 years ago[PARISC] Fix unwinder on 64-bit kernels
Randolph Chung [Tue, 12 Jun 2007 06:27:32 +0000 (14:27 +0800)]
[PARISC] Fix unwinder on 64-bit kernels

The unwinder was broken by the shift of PAGE_OFFSET in order to increase the
size of the vmalloc area on 64-bit kernels.

Signed-off-by: Randolph Chung <tausq@debian.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
17 years ago[PARISC] Handle wrapping in expand_upwards()
Helge Deller [Tue, 19 Dec 2006 18:28:33 +0000 (19:28 +0100)]
[PARISC] Handle wrapping in expand_upwards()

Function expand_upwards() did not guarded against wrapping
around to address 0. This fixes the adjtimex02 testcase from
the Linux Test Project on a 32bit PARISC kernel.

[expand_upwards is only used on parisc and ia64; it looks like it does
 the right thing on both. --kyle]

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: Tony Luck <tony.luck@intel.com>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
17 years ago[PARISC] stop lcd driver from stripping initial whitespace
Julian Stecklina [Mon, 18 Jun 2007 17:56:28 +0000 (19:56 +0200)]
[PARISC] stop lcd driver from stripping initial whitespace

Signed-off-by: Julian Stecklina <der_julian@web.de>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
17 years ago[AGPGART] intel_agp: don't load if no IGD and AGP port
Wang Zhenyu [Thu, 21 Jun 2007 05:43:18 +0000 (13:43 +0800)]
[AGPGART] intel_agp: don't load if no IGD and AGP port

After i915 chip, GMCH has no AGP port. Origin bridge driver in device
table will try to access illegal regs like APBASE, APSIZE, etc. This
may cause problem.

So mark them as NULL in the table, we won't load if no IGD got detect
and bridge has no AGP port.

Signed-off-by: Wang Zhenyu <zhenyu.z.wang@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>
17 years agoMerge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Thu, 21 Jun 2007 00:11:44 +0000 (17:11 -0700)]
Merge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  fix module_param mistake in it821x
  ahci: fix PORTS_IMPL override
  kerneldoc fix in libata
  libata: more NONCQ devices
  pata_it821x: (partially) fix DMA in RAID mode
  PATA: Add the MCP73/77 support to PATA driver

17 years agofix module_param mistake in it821x
Stas Sergeev [Wed, 20 Jun 2007 18:42:13 +0000 (22:42 +0400)]
fix module_param mistake in it821x

The attached patch fixes a trivial
mistake in a MODULE_PARAM_DESC of pata_it821x
driver. The parameter name in MODULE_PARAM_DESC
should match the one in module_param_named.

Signed-off-by: Stas Sergeev <stsp@aknet.ru>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoahci: fix PORTS_IMPL override
Tejun Heo [Tue, 19 Jun 2007 09:52:56 +0000 (18:52 +0900)]
ahci: fix PORTS_IMPL override

If PORTS_IMPL register is zero, ahci initialize it to full mask
corresponding to nr_ports in the CAP register.  hpriv->cap, which is
initialized at the end of the function, is incorrectly used as value
of CAP causing ahci to always override PORTS_IMPL to 0x1 if it's zero.
Fix it.

This fixes a bug where early ich6 ahci can only access the first port.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agokerneldoc fix in libata
Henrik Kretzschmar [Tue, 19 Jun 2007 08:10:50 +0000 (10:10 +0200)]
kerneldoc fix in libata

Fix parameter name from ata_dev_reread_id() in libata-core.c for kerneldoc.

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agolibata: more NONCQ devices
Tejun Heo [Mon, 18 Jun 2007 05:38:41 +0000 (14:38 +0900)]
libata: more NONCQ devices

More for the NCQ blacklist.  One hitachi and one raptor.  Other
members of these families of drives are already on the list, so no
surprises.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agopata_it821x: (partially) fix DMA in RAID mode
Bartlomiej Zolnierkiewicz [Mon, 11 Jun 2007 09:40:07 +0000 (11:40 +0200)]
pata_it821x: (partially) fix DMA in RAID mode

Code intended to check DMA status was checking DMA command register.

Moreover firmware seems to "forget" to set DMA capable bit for the
slave device (at least in RAID mode but without ITE RAID volumes) so
check device ID for DMA capable bit when deciding whether to use DMA
and remove DMA status check completely.

Thanks to Pavol Simo for the bugreport and testing the initial fix.

This change unfortunately still doesn't fix DMA in RAID mode (which
works fine with IDE it821x) but Alan is working on the missing pieces
(pata_it821x vs libata EH issues).

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Acked-by: Alan Cox <alan@redhat.com>
Cc: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoPATA: Add the MCP73/77 support to PATA driver
Peer Chen [Thu, 7 Jun 2007 10:23:12 +0000 (18:23 +0800)]
PATA: Add the MCP73/77 support to PATA driver

Add the MCP73/MCP77 support to PATA driver.
The patch base on kernel 2.6.22-rc4

Signed-off-by: Peer Chen <peerchen@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agocxgb3 - MAC watchdog update
Divy Le Ray [Wed, 30 May 2007 17:01:50 +0000 (10:01 -0700)]
cxgb3 - MAC watchdog update

Fix variables initialization and usage in the MAC watchdog.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agocxgb3 - Stop mac RX when changing MTU
Divy Le Ray [Wed, 30 May 2007 17:01:44 +0000 (10:01 -0700)]
cxgb3 - Stop mac RX when changing MTU

Rx traffic needs to be halted when the MTU is changed
to avoid a potential chip hang.
Reset/restore MAC filters around a MTU change.
Also fix the pause frames high materwark setting.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agocxgb3 - Fix direct XAUI support
Divy Le Ray [Wed, 30 May 2007 17:01:39 +0000 (10:01 -0700)]
cxgb3 - Fix direct XAUI support

Check all lanes for link status on direct XAUI cards.
Don't assume that direct XAUI always uses XGMAC 1.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agocxgb3 - fix netpoll hanlder
Divy Le Ray [Wed, 30 May 2007 17:01:34 +0000 (10:01 -0700)]
cxgb3 - fix netpoll hanlder

Fix netpoll handler to work with line interrupt, msi and msi-x.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agocxgb3 - fix skb->dev dereference
Divy Le Ray [Wed, 30 May 2007 17:01:29 +0000 (10:01 -0700)]
cxgb3 - fix skb->dev dereference

eth_type_trans() now sets skb->dev.
References to skb->dev should happen after it is called.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agonatsemi irq flags
Gregory Haskins [Fri, 15 Jun 2007 21:17:50 +0000 (17:17 -0400)]
natsemi irq flags

The spinlock irq flags should be a unsigned long to properly support 64 bit

Signed-off-by: Gregory Haskins <ghaskins@novell.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoforcedeth: use unicast receive mode for WoL
Tim Mann [Thu, 14 Jun 2007 20:16:38 +0000 (13:16 -0700)]
forcedeth: use unicast receive mode for WoL

I happened to notice that a system with an NVidia NIC using the
forcedeth driver won't wake-on-LAN if the interface was in promiscuous
mode when you power off.  By experiment, it looks like
the hardware needs to have NvRegPacketFilterFlags set to
NVREG_PFF_ALWAYS|NVREG_PFF_MYADDR (i.e., receive unicast packets to my
address) in order for WoL to work.

Jeff Garzik writes: "NVIDIA says the patch looks OK."  I didn't venture
to insert a signed-off-by line with his name on it, though.

Signed-off-by: Tim Mann <mann@vmware.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: netiucv inlining cleanup
Martin Schwidefsky [Wed, 20 Jun 2007 11:03:57 +0000 (13:03 +0200)]
s390: netiucv inlining cleanup

The recent iucv rework patches re-introduced some unnecessary inlines.
Remove them again.

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: netiucv spinlock initializer cleanup
Thomas Gleixner [Wed, 20 Jun 2007 11:02:55 +0000 (13:02 +0200)]
s390: netiucv spinlock initializer cleanup

spinlock initializer cleanup in netiucv.c

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: don't call iucv_path_connect from tasklet context
Ursula Braun [Wed, 20 Jun 2007 11:01:30 +0000 (13:01 +0200)]
s390: don't call iucv_path_connect from tasklet context

net/iucv/iucv.c creates the requirement for
iucv_path_connect not to be called from tasklet context anymore.
An extra checking is added in case of a failing netiucv_tx
to fulfil this requirement for netiucv.

Signed-off-by: Ursula Braun <braunu@de.ibm.com>
Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: Use ccw_device_get_id() in qeth/claw drivers
Cornelia Huck [Wed, 20 Jun 2007 11:00:20 +0000 (13:00 +0200)]
s390: Use ccw_device_get_id() in qeth/claw drivers

Use ccw_device_get_id() to get a device number
instead of parsing the ccw device's bus id.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: qeth: wrong packet length in qdio header
Frank Pavlic [Wed, 20 Jun 2007 10:59:14 +0000 (12:59 +0200)]
s390: qeth: wrong packet length in qdio header

Packets Length in qdio header is broken when using
EDDP on Layer2 devices. This leads to skb_under_panic on receiver
system when running on z/VM GuestLAN devices.

Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: avoid inconsistent lock state in qeth
Ursula Braun [Wed, 20 Jun 2007 10:58:02 +0000 (12:58 +0200)]
s390: avoid inconsistent lock state in qeth

ipv6_regen_rndid in net/ipv6/addrconf.c makes use of "write_lock_bh"
for its inet6_dev->lock. It may run in softirq-context.
qeth makes use of "read_lock" for the same inet6_dev->lock.
To avoid a potential deadlock situation, qeth should make use of
"read_lock_bh" for its usages of inet6_dev->lock.

Signed-off-by: Ursula Braun <braunu@de.ibm.com>
Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: qeth driver does not recover
Ursula Braun [Wed, 20 Jun 2007 10:56:49 +0000 (12:56 +0200)]
s390: qeth driver does not recover

While first recovery continues, the card issues
a STARTLAN command itself. In this case qeth
schedules another recovery. This second
recovery is cancelled because of an already running first recovery.
Stop first recovery in case of 0xe080.

Signed-off-by: Ursula Braun <braunu@de.ibm.com>
Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agos390: print correct level for HiperSockets devices
Ursula Braun [Wed, 20 Jun 2007 10:55:18 +0000 (12:55 +0200)]
s390: print correct level for HiperSockets devices

For real HiperSockets the EBCDIC-ASCII conversion is not necessary.
This is only needed for z/VM GuestLAN devices.

Signed-off-by: Ursula Braun <braunu@de.ibm.com>
Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agobonding: Fix 802.3ad no carrier on "no partner found" instance
Jay Vosburgh [Thu, 14 Jun 2007 05:11:34 +0000 (22:11 -0700)]
bonding: Fix 802.3ad no carrier on "no partner found" instance

Modify carrier state determination for 802.3ad mode to comply
with section 43.3.9 of IEEE 802.3, which requires that "Links that are
not successful candidates for aggregation (e.g., links that are attached
to other devices that cannot perform aggregation or links that have been
manually configured to be non-aggregatable) are enabled to operate as
individual IEEE 802.3 links."

Bug reported by Laurent Chavey <chavey@google.com>.  This patch
is an updated version of his patch that changes the wording of
commentary and adds an update to the driver version.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Laurent Chavey <chavey@google.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agobonding: Fix use after free in unregister path
Jay Vosburgh [Tue, 19 Jun 2007 18:12:12 +0000 (11:12 -0700)]
bonding: Fix use after free in unregister path

The following patch (based on a patch from Stephen Hemminger
<shemminger@linux-foundation.org>) removes use after free conditions in
the unregister path for the bonding master.  Without this patch, an
operation of the form "echo -bond0 > /sys/class/net/bonding_masters"
would trigger a NULL pointer dereference in sysfs.  I was not able to
induce the failure with the non-sysfs code path, but for consistency I
updated that code as well.

I also did some testing of the bonding /proc file being open
while the bond is being deleted, and didn't see any problems there.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agospidernet: checksum and ethtool
Stephen Hemminger [Mon, 11 Jun 2007 18:35:23 +0000 (13:35 -0500)]
spidernet: checksum and ethtool

It doesn't look like spidernet hardware can really checksum all protocols,
the code looks like it does IPV4 only.  If so, it should use NETIF_F_IP_CSUM
instead of NETIF_F_HW_CSUM.

The driver doesn't need it's own get/set for ethtool tx csum, and it
should use the standard ethtool_op_get_link.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agospidernet: turn off descriptor chain end interrupt.
Linas Vepstas [Mon, 11 Jun 2007 18:29:03 +0000 (13:29 -0500)]
spidernet: turn off descriptor chain end interrupt.

At some point, the transmit descriptor chain end interrupt (TXDCEINT)
was turned on. This is a mistake; and it damages small packet
transmit performance, as it results in a huge storm of interrupts.
Turn it off.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agospidernet: silence the ramfull messages
Linas Vepstas [Mon, 11 Jun 2007 18:26:50 +0000 (13:26 -0500)]
spidernet: silence the ramfull messages

Although the previous patch resolved issues with hangs when the
RX ram full interrupt is encountered, there are still situations
where lots of RX ramfull interrupts arrive, resulting in a noisy
log in syslog. There is no need for this.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agospidernet: Don't terminate the RX ring
Linas Vepstas [Mon, 11 Jun 2007 18:23:42 +0000 (13:23 -0500)]
spidernet: Don't terminate the RX ring

The terminated RX ring will cause trouble during the RX ram full
conditions, leading to a hung driver, as the hardware can't find
the next descr.  There is no real reason to terminate the RX ring;
it doesn't make the operation any smooother, and it does
require an extra sync. So don't do it.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agospidernet: Cure RX ram full bug
Linas Vepstas [Mon, 11 Jun 2007 18:21:13 +0000 (13:21 -0500)]
spidernet: Cure RX ram full bug

This patch fixes a rare deadlock that can occur when the kernel
is not able to empty out the RX ring quickly enough. Below follows
a detailed description of the bug and the fix.

As long as the OS can empty out the RX buffers at a rate faster than
the hardware can fill them, there is no problem. If, for some reason,
the OS fails to empty the RX ring fast enough, the hardware GDACTDPA
pointer will catch up to the head, notice the not-empty condition,
ad stop. However, RX packets may still continue arriving on the wire.
The spidernet chip can save some limited number of these in local RAM.
When this local ram fills up, the spider chip will issue an interrupt
indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit
will be set in GHIINT1STS).  When te RX ram full condition occurs,
a certain bug/feature is triggered that has to be specially handled.
This section describes the special handling for this condition.

When the OS finally has a chance to run, it will empty out the RX ring.
In particular, it will clear the descriptor on which the hardware had
stopped. However, once the hardware has decided that a certain
descriptor is invalid, it will not restart at that descriptor; instead
it will restart at the next descr. This potentially will lead to a
deadlock condition, as the tail pointer will be pointing at this descr,
which, from the OS point of view, is empty; the OS will be waiting for
this descr to be filled. However, the hardware has skipped this descr,
and is filling the next descrs. Since the OS doesn't see this, there
is a potential deadlock, with the OS waiting for one descr to fill,
while the hardware is waiting for a differen set of descrs to become
empty.

A call to show_rx_chain() at this point indicates the nature of the
problem. A typical print when the network is hung shows the following:

net eth1: Spider RX RAM full, incoming packets might be discarded!
net eth1: Total number of descrs=256
net eth1: Chain tail located at descr=255
net eth1: Chain head is at 255
net eth1: HW curr desc (GDACTDPA) is at 0
net eth1: Have 1 descrs with stat=xa0800000
net eth1: HW next desc (GDACNEXTDA) is at 1
net eth1: Have 127 descrs with stat=x40800101
net eth1: Have 1 descrs with stat=x40800001
net eth1: Have 126 descrs with stat=x40800101
net eth1: Last 1 descrs with stat=xa0800000

Both the tail and head pointers are pointing at descr 255, which is
marked xa... which is "empty". Thus, from the OS point of view, there
is nothing to be done. In particular, there is the implicit assumption
that everything in front of the "empty" descr must surely also be empty,
as explained in the last section. The OS is waiting for descr 255 to
become non-empty, which, in this case, will never happen.

The HW pointer is at descr 0. This descr is marked 0x4.. or "full".
Since its already full, the hardware can do nothing more, and thus has
halted processing. Notice that descrs 0 through 254 are all marked
"full", while descr 254 and 255 are empty. (The "Last 1 descrs" is
descr 254, since tail was at 255.) Thus, the system is deadlocked,
and there can be no forward progress; the OS thinks there's nothing
to do, and the hardware has nowhere to put incoming data.

This bug/feature is worked around with the spider_net_resync_head_ptr()
routine. When the driver receives RX interrupts, but an examination
of the RX chain seems to show it is empty, then it is probable that
the hardware has skipped a descr or two (sometimes dozens under heavy
network conditions). The spider_net_resync_head_ptr() subroutine will
search the ring for the next full descr, and the driver will resume
operations there.  Since this will leave "holes" in the ring, there
is also a spider_net_resync_tail_ptr() that will skip over such holes.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agospidernet: null out skb pointer after its been used.
Linas Vepstas [Mon, 11 Jun 2007 18:17:02 +0000 (13:17 -0500)]
spidernet: null out skb pointer after its been used.

Avoid kernel crash in mm/slab.c due to double-free of pointer.

If the ethernet interface is brought down while there is still
RX traffic in flight, the device shutdown routine can end up
trying to double-free an skb, leading to a crash in mm/slab.c
Avoid the double-free by nulling out the skb pointer.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394...
Linus Torvalds [Wed, 20 Jun 2007 22:39:50 +0000 (15:39 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
  firewire: Only set client->iso_context if allocation was successful.
  ieee1394: fix to ether1394_tx in ether1394.c
  firewire: fix hang after card ejection

17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Wed, 20 Jun 2007 22:39:24 +0000 (15:39 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IB/mlx4: Make sure inline data segments don't cross a 64 byte boundary
  IB/mlx4: Handle FW command interface rev 3
  IB/mlx4: Handle buffer wraparound in __mlx4_ib_cq_clean()
  IB/mlx4: Get rid of max_inline_data calculation
  IB/mlx4: Handle new FW requirement for send request prefetching
  IB/mlx4: Fix warning in rounding up queue sizes
  IB/mlx4: Fix handling of wq->tail for send completions

17 years agofirewire: Only set client->iso_context if allocation was successful.
Kristian Høgsberg [Wed, 20 Jun 2007 21:48:07 +0000 (17:48 -0400)]
firewire: Only set client->iso_context if allocation was successful.

This patch fixes an OOPS on cdev release for an fd where iso context
creation failed.

Signed-off-by: Kristian Høgsberg <krh@redhat.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
17 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Wed, 20 Jun 2007 21:40:52 +0000 (14:40 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] Don't drag a platform specific header into generic arch code.

17 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Wed, 20 Jun 2007 21:28:54 +0000 (14:28 -0700)]
Merge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc:
  [POWERPC] Fix powermac late initcall to only run on powermac
  [POWERPC] PowerPC: Prevent data exception in kernel space (32-bit)

17 years agoFix up CREDIT entry ordering
Li Yang [Wed, 20 Jun 2007 12:23:34 +0000 (20:23 +0800)]
Fix up CREDIT entry ordering

Reorder my CREDIT entry to make it in alphabetic order by last name.

Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: fix link warning between for .text and .init.text
Yinghai Lu [Wed, 20 Jun 2007 10:23:39 +0000 (12:23 +0200)]
x86_64: fix link warning between for .text and .init.text

WARNING: arch/x86_64/kernel/built-in.o(.text+0xace9): Section mismatch: reference to .init.text: (between 'get_mtrr_state' and 'mtrr_wrmsr')
WARNING: arch/x86_64/kernel/built-in.o(.text+0xad09): Section mismatch: reference to .init.text: (between 'get_mtrr_state' and 'mtrr_wrmsr')
WARNING: arch/x86_64/kernel/built-in.o(.text+0xad38): Section mismatch: reference to .init.text: (between 'get_mtrr_state' and 'mtrr_wrmsr')
WARNING: drivers/built-in.o(.text+0x3a680): Section mismatch: reference to .init.text:acpi_map_pxm_to_node (between 'acpi_get_node' and 'acpi_lock_ac_dir')

AK: also marked mtrr_bp_init __init to avoid some more warnings

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Acked-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86: change_page_attr bandaids
Andi Kleen [Wed, 20 Jun 2007 10:23:36 +0000 (12:23 +0200)]
x86: change_page_attr bandaids

- Disable CLFLUSH again; it is still broken. Always do WBINVD.
- Always flush in the i386 case, not only when there are deferred pages.

These are both brute-force inefficient fixes, to be improved
next release cycle.

The changes to i386 are a little more extensive than strictly
needed (some dead code added), but it is more similar to the x86-64 version
now and the dead code will be used soon.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86: Disable KPROBES with DEBUG_RODATA for now
Andi Kleen [Wed, 20 Jun 2007 10:23:35 +0000 (12:23 +0200)]
x86: Disable KPROBES with DEBUG_RODATA for now

Right now Kprobes cannot write to the write protected kernel text when
DEBUG_RODATA is enabled. Disallow this in Kconfig for now.

Temporary fix for 2.6.22. In .23 add code to temporarily
unprotect it.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86: Only make Macintosh drivers default on Macs
Andi Kleen [Wed, 20 Jun 2007 10:23:34 +0000 (12:23 +0200)]
x86: Only make Macintosh drivers default on Macs

It's already annoying that they appear on x86 now -- that's for the 3button
emulation needed on x86 macs -- but at least don't make them default.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: Quieten Atari keyboard warnings in Kconfig
Andi Kleen [Wed, 20 Jun 2007 10:23:33 +0000 (12:23 +0200)]
x86_64: Quieten Atari keyboard warnings in Kconfig

Not directly related to x86, but I got tired of seeing these warnings on every
kconfig update when building on a non m68k box:

drivers/input/keyboard/Kconfig:170:warning: 'select' used by config symbol 'KEYBOARD_ATARI' refers to undefined symbol 'ATARI_KBD_CORE'
drivers/input/mouse/Kconfig:182:warning: 'select' used by config symbol 'MOUSE_ATARI' refers to undefined symbol 'ATARI_KBD_CORE'

I moved the definition of ATARI_KBD_CORE into drivers/input/keyboard/Kconfig
so it's always seen by Kconfig.

Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86: Disable DAC on VIA bridges
Andi Kleen [Wed, 20 Jun 2007 10:23:32 +0000 (12:23 +0200)]
x86: Disable DAC on VIA bridges

Several reports that VIA bridges don't support DAC and corrupt
data.  I don't know if it's fixed, but let's just blacklist
them all for now.

It can be overwritten with iommu=usedac

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: Fix eventd/timerfd syscalls
Andi Kleen [Wed, 20 Jun 2007 10:23:31 +0000 (12:23 +0200)]
x86_64: Fix eventd/timerfd syscalls

They had the same syscall number.

Pointed out by Davide Libenzi

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: Fix readahead/sync_file_range/fadvise64 compat calls
Andi Kleen [Wed, 20 Jun 2007 10:23:30 +0000 (12:23 +0200)]
x86_64: Fix readahead/sync_file_range/fadvise64 compat calls

Correctly convert the u64 arguments from 32bit to 64bit.

Pointed out by Heiko Carstens.

I guess this proves Linus' theory that nobody uses the more exotic Linux
specific syscalls.  It wasn't discovered by a user.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[MIPS] Don't drag a platform specific header into generic arch code.
Ralf Baechle [Wed, 20 Jun 2007 21:27:10 +0000 (22:27 +0100)]
[MIPS] Don't drag a platform specific header into generic arch code.

For some platforms it's definitions may conflict.  So that's the one-liner.
The rest is 10 square kilometers of collateral damage fixup this include
used to paper over.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
17 years ago[POWERPC] Fix powermac late initcall to only run on powermac
Tony Breeds [Wed, 20 Jun 2007 05:17:29 +0000 (15:17 +1000)]
[POWERPC] Fix powermac late initcall to only run on powermac

Current ppc64_defconfig kernel fails to boot on iSeries, dying with:

Unable to handle kernel paging request for data at address 0x00000000
Faulting instruction address: 0xc00000000071b258
Oops: Kernel access of bad area, sig: 11 [#1]
SMP NR_CPUS=32 iSeries
<snip>
NIP [c00000000071b258] .iSeries_src_init+0x34/0x64
LR [c000000000701bb4] .kernel_init+0x1fc/0x3bc
Call Trace:
[c000000007d0be30] [0000000000008000] 0x8000 (unreliable)
[c000000007d0bea0] [c000000000701bb4] .kernel_init+0x1fc/0x3bc
[c000000007d0bf90] [c0000000000262d4] .kernel_thread+0x4c/0x68
Instruction dump:
e922cba8 3880ffff 78840420 f8010010 f821ff91 60000000 e8090000 78095fe3
4182002c e922cb58 e862cbb0 e9290140 <e8090000f8410028 7c0903a6 e9690010
Kernel panic - not syncing: Attempted to kill init!

This happens because some powermac code unconditionally sets
ppc_md.progress to NULL.  This patch makes sure the powermac late
initcall is only run on powermac machines.

Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] PowerPC: Prevent data exception in kernel space (32-bit)
Segher Boessenkool [Tue, 19 Jun 2007 15:07:04 +0000 (01:07 +1000)]
[POWERPC] PowerPC: Prevent data exception in kernel space (32-bit)

The "is_exec" branch of the protection check in do_page_fault()
didn't do anything on 32-bit PowerPC.  So if a userland program
jumps to a page with Linux protection flags "---p", all the tests
happily fall through, and handle_mm_fault() is called, which in
turn calls handle_pte_fault(), which calls update_mmu_cache(),
which goes flush the dcache to a page with no access rights.

Boom.

This fixes it.

Signed-off-by: Segher Boessenkool <segher@kernel.crashing.org>
Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] rheap - eliminates internal fragments caused by alignment
Li Yang [Mon, 18 Jun 2007 11:29:21 +0000 (19:29 +0800)]
[POWERPC] rheap - eliminates internal fragments caused by alignment

The patch adds fragments caused by rh_alloc_align() back to free list, instead
of allocating the whole chunk of memory.  This will greatly improve memory
utilization managed by rheap.

It solves MURAM not enough problem with 3 UCCs enabled on MPC8323.

Signed-off-by: Li Yang <leoli@freescale.com>
Acked-by: Joakim Tjernlund <joakim.tjernlund@transmode.se>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh64-2.6
Linus Torvalds [Tue, 19 Jun 2007 15:07:52 +0000 (08:07 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh64-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh64-2.6:
  sh64: Handle -ERESTART_RESTARTBLOCK for restartable syscalls.

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh-2.6
Linus Torvalds [Tue, 19 Jun 2007 15:07:34 +0000 (08:07 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh-2.6:
  sh: Handle -ERESTART_RESTARTBLOCK for restartable syscalls.
  sh: oops_enter()/oops_exit() in die().
  sh: Fix restartable syscall arg5 clobbering.

17 years agoMerge branch 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Tue, 19 Jun 2007 15:07:08 +0000 (08:07 -0700)]
Merge branch 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6

* 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [S390] Move psw_set_key.
  [S390] Add oops_enter()/oops_exit() calls to die().
  [S390] Print list of modules on die().
  [S390] Fix yet another two section mismatches.
  [S390] Fix zfcpdump header
  [S390] Missing blank when appending cio_ignore kernel parameter

17 years agoMerge branch 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-2.6
Linus Torvalds [Tue, 19 Jun 2007 15:06:25 +0000 (08:06 -0700)]
Merge branch 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-2.6

* 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-2.6:
  [XFS] Update the MAINTAINERS file entry for XFS - change git repo name.
  [XFS] s/memclear_highpage_flush/zero_user_page/
  [XFS] Update the MAINTAINERS file entry for XFS.

17 years ago[S390] Move psw_set_key.
Heiko Carstens [Tue, 19 Jun 2007 11:10:06 +0000 (13:10 +0200)]
[S390] Move psw_set_key.

Move psw_set_key() from ptrace.h to processor.h which is a more
suitable place for it. In addition the moves makes the function
invisible to user space.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
17 years ago[S390] Add oops_enter()/oops_exit() calls to die().
Heiko Carstens [Tue, 19 Jun 2007 11:10:05 +0000 (13:10 +0200)]
[S390] Add oops_enter()/oops_exit() calls to die().

This is mainly to switch off all potentially debugging stuff that
won't report anything useful after an oops happened.
Besided that setting pause_on_oops will work too, but doesn't make
too much sense on s390.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
17 years ago[S390] Print list of modules on die().
Heiko Carstens [Tue, 19 Jun 2007 11:10:04 +0000 (13:10 +0200)]
[S390] Print list of modules on die().

Print list of modules on die() like a lot of other architectures do.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
17 years ago[S390] Fix yet another two section mismatches.
Heiko Carstens [Tue, 19 Jun 2007 11:10:03 +0000 (13:10 +0200)]
[S390] Fix yet another two section mismatches.

WARNING: arch/s390/kernel/built-in.o(.text+0xb92a):
 Section mismatch: reference to .init.text:start_secondary
 (between 'restart_addr' and 'stack_overflow')
WARNING: arch/s390/appldata/built-in.o(.data+0xdc):
 Section mismatch: reference to .init.text:
 (between 'appldata_nb' and 'appldata_timer_lock')

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
17 years ago[S390] Fix zfcpdump header
Michael Holzheu [Tue, 19 Jun 2007 11:10:02 +0000 (13:10 +0200)]
[S390] Fix zfcpdump header

Added members for volume number and real memory size to header information.

Signed-off-by: Michael Holzheu <holzheu@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
17 years ago[S390] Missing blank when appending cio_ignore kernel parameter
Michael Holzheu [Tue, 19 Jun 2007 11:10:01 +0000 (13:10 +0200)]
[S390] Missing blank when appending cio_ignore kernel parameter

When appending the 'cio_ignore' kernel parameter to the command line, a blank
has to be inserted in order to separate 'cio_ignore' from the preceding kernel
parameters.

Signed-off-by: Michael Holzheu <holzheu@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
17 years ago[XFS] Update the MAINTAINERS file entry for XFS - change git repo name.
Tim Shimmin [Tue, 19 Jun 2007 05:26:35 +0000 (15:26 +1000)]
[XFS] Update the MAINTAINERS file entry for XFS - change git repo name.

Make the git repository bare and so give it the conventional .git suffix.

Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years ago[XFS] s/memclear_highpage_flush/zero_user_page/
Christoph Hellwig [Thu, 24 May 2007 05:27:17 +0000 (15:27 +1000)]
[XFS] s/memclear_highpage_flush/zero_user_page/

SGI-PV: 957103
SGI-Modid: xfs-linux-melb:xfs-kern:28678a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
17 years agoMerge branch 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-2.6 into for-linus
Tim Shimmin [Tue, 19 Jun 2007 05:06:15 +0000 (15:06 +1000)]
Merge branch 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-2.6 into for-linus

17 years ago[POWERPC] Fix snd-powermac refcounting bugs
Benjamin Herrenschmidt [Tue, 19 Jun 2007 04:37:39 +0000 (14:37 +1000)]
[POWERPC] Fix snd-powermac refcounting bugs

The old snd-powermac driver has some serious refcounting issues when
initialisation fails, which is the case on all new machines with
a layout-id since those are handled by the new snd-aoa driver.

Some of those bugs seem to have been under the radar for some time
(like double pci_dev_put), but one was actually added in 2.6.22 with
Stephen attempt at teaching refcounting to the driver which didn't
do it at all.

This patch fixes both, thus removing all sort of kref errors that
would happen if that driver gets loaded on a G5 machine or a recent
PowerBook due to OF nodes left around with a 0 refcount.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years agosh64: Handle -ERESTART_RESTARTBLOCK for restartable syscalls.
Paul Mundt [Tue, 19 Jun 2007 03:41:32 +0000 (12:41 +0900)]
sh64: Handle -ERESTART_RESTARTBLOCK for restartable syscalls.

The current implementation only handles -ERESTARTNOHAND, whereas we
also need to handle -ERESTART_RESTARTBLOCK in the handle_signal()
case for restartable system calls. Follows the sh change.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
17 years agosh: Handle -ERESTART_RESTARTBLOCK for restartable syscalls.
Paul Mundt [Tue, 19 Jun 2007 03:33:21 +0000 (12:33 +0900)]
sh: Handle -ERESTART_RESTARTBLOCK for restartable syscalls.

The current implementation only handles -ERESTARTNOHAND, whereas we
also need to handle -ERESTART_RESTARTBLOCK in the handle_signal()
case for restartable system calls.

As noted by Carl:

This fixes the LTP test nanosleep03 - the current kernel causes
-ERESTART_RESTARTBLOCK to reach user space rather than the correct
-EINTR.

Reported-by: Carl Shaw <shaw.carl@gmail.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
17 years agoFix possible runqueue lock starvation in wait_task_inactive()
Linus Torvalds [Mon, 18 Jun 2007 16:34:40 +0000 (09:34 -0700)]
Fix possible runqueue lock starvation in wait_task_inactive()

Miklos Szeredi reported very long pauses (several seconds, sometimes
more) on his T60 (with a Core2Duo) which he managed to track down to
wait_task_inactive()'s open-coded busy-loop.

He observed that an interrupt on one core tries to acquire the
runqueue-lock but does not succeed in doing so for a very long time -
while wait_task_inactive() on the other core loops waiting for the first
core to deschedule a task (which it wont do while spinning in an
interrupt handler).

This rewrites wait_task_inactive() to do all its waiting optimistically
without any locks taken at all, and then just double-check the end
result with the proper runqueue lock held over just a very short
section.  If there were races in the optimistic wait, of a preemption
event scheduled the process away, we simply re-synchronize, and start
over.

So the code now looks like this:

repeat:
/* Unlocked, optimistic looping! */
rq = task_rq(p);
while (task_running(rq, p))
cpu_relax();

/* Get the *real* values */
rq = task_rq_lock(p, &flags);
running = task_running(rq, p);
array = p->array;
task_rq_unlock(rq, &flags);

/* Check them.. */
if (unlikely(running)) {
cpu_relax();
goto repeat;
}

/* Preempted away? Yield if so.. */
if (unlikely(array)) {
yield();
goto repeat;
}

Basically, that first "while()" loop is done entirely without any
locking at all (and doesn't check for the case where the target process
might have been preempted away), and so it's possibly "incorrect", but
we don't really care.  Both the runqueue used, and the "task_running()"
check might be the wrong tests, but they won't oops - they just mean
that we could possibly get the wrong results due to lack of locking and
exit the loop early in the case of a race condition.

So once we've exited the loop, we then get the proper (and careful) rq
lock, and check the running/runnable state _safely_.  And if it turns
out that our quick-and-dirty and unsafe loop was wrong after all, we
just go back and try it all again.

(The patch also adds a lot of comments, which is the actual bulk of it
all, to make it more obvious why we can do these things without holding
the locks).

Thanks to Miklos for all the testing and tracking it down.

Tested-by: Miklos Szeredi <miklos@szeredi.hu>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosched: fix SysRq-N (normalize RT tasks)
Ingo Molnar [Sun, 17 Jun 2007 16:37:45 +0000 (18:37 +0200)]
sched: fix SysRq-N (normalize RT tasks)

Gene Heskett reported the following problem while testing CFS: SysRq-N
is not always effective in normalizing tasks back to SCHED_OTHER.

The reason for that turns out to be the following bug:

 - normalize_rt_tasks() uses for_each_process() to iterate through all
   tasks in the system.  The problem is, this method does not iterate
   through all tasks, it iterates through all thread groups.

The proper mechanism to enumerate over all threads is to use a
do_each_thread() + while_each_thread() loop.

Reported-by: Gene Heskett <gene.heskett@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6
Linus Torvalds [Mon, 18 Jun 2007 17:38:09 +0000 (10:38 -0700)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6:
  [SCSI] ESP: Don't forget to clear ESP_FLAG_RESETTING.
  [SCSI] fusion: fix for BZ 8426 - massive slowdown on SCSI CD/DVD drive

17 years agoFix signalfd interaction with thread-private signals
Benjamin Herrenschmidt [Mon, 11 Jun 2007 22:16:18 +0000 (08:16 +1000)]
Fix signalfd interaction with thread-private signals

Don't let signalfd dequeue private signals off other threads (in the
case of things like SIGILL or SIGSEGV, trying to do so would result
in undefined behaviour on who actually gets the signal, since they
are force unblocked).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoRevert "futex_requeue_pi optimization"
Thomas Gleixner [Sun, 17 Jun 2007 19:11:10 +0000 (21:11 +0200)]
Revert "futex_requeue_pi optimization"

This reverts commit d0aa7a70bf03b9de9e995ab272293be1f7937822.

It not only introduced user space visible changes to the futex syscall,
it is also non-functional and there is no way to fix it proper before
the 2.6.22 release.

The breakage report ( http://lkml.org/lkml/2007/5/12/17 ) went
unanswered, and unfortunately it turned out that the concept is not
feasible at all.  It violates the rtmutex semantics badly by introducing
a virtual owner, which hacks around the coupling of the user-space
pi_futex and the kernel internal rt_mutex representation.

At the moment the only safe option is to remove it fully as it contains
user-space visible changes to broken kernel code, which we do not want
to expose in the 2.6.22 release.

The patch reverts the original patch mostly 1:1, but contains a couple
of trivial manual cleanups which were necessary due to patches, which
touched the same area of code later.

Verified against the glibc tests and my own PI futex tests.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Ulrich Drepper <drepper@redhat.com>
Cc: Pierre Peiffer <pierre.peiffer@bull.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoIB/mlx4: Make sure inline data segments don't cross a 64 byte boundary
Roland Dreier [Mon, 18 Jun 2007 16:23:47 +0000 (09:23 -0700)]
IB/mlx4: Make sure inline data segments don't cross a 64 byte boundary

Inline data segments in send WQEs are not allowed to cross a 64 byte
boundary.  We use inline data segments to hold the UD headers for MLX
QPs (QP0 and QP1).  A send with GRH on QP1 will have a UD header that
is too big to fit in a single inline data segment without crossing a
64 byte boundary, so split the header into two inline data segments.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoIB/mlx4: Handle FW command interface rev 3
Roland Dreier [Mon, 18 Jun 2007 15:15:02 +0000 (08:15 -0700)]
IB/mlx4: Handle FW command interface rev 3

Upcoming firmware introduces command interface revision 3, which
changes the way port capabilities are queried and set.  Update the
driver to handle both the new and old command interfaces by adding a
new MLX4_FLAG_OLD_PORT_CMDS that it is set after querying the firmware
interface revision and then using the correct interface based on the
setting of the flag.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoIB/mlx4: Handle buffer wraparound in __mlx4_ib_cq_clean()
Jack Morgenstein [Mon, 18 Jun 2007 15:13:59 +0000 (08:13 -0700)]
IB/mlx4: Handle buffer wraparound in __mlx4_ib_cq_clean()

When compacting CQ entries, we need to set the correct value of the
ownership bit in case the value is different between the index we copy
the CQE from and the index we copy it to.

Found by Ronni Zimmerman of Mellanox.

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoIB/mlx4: Get rid of max_inline_data calculation
Roland Dreier [Mon, 18 Jun 2007 15:13:53 +0000 (08:13 -0700)]
IB/mlx4: Get rid of max_inline_data calculation

The calculation of max_inline_data in set_kernel_sq_size() is bogus,
since it doesn't take into account the fact that inline segments may
not cross a 64-byte boundary, and hence multiple inline segments will
probably need to be used to post large inline sends.

We don't support inline sends for kernel QPs anyway, so there's no
point in doing this calculation anyway, since the field is just zeroed
out a little later.  So just delete the bogus calculation.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoIB/mlx4: Handle new FW requirement for send request prefetching
Roland Dreier [Mon, 18 Jun 2007 15:13:48 +0000 (08:13 -0700)]
IB/mlx4: Handle new FW requirement for send request prefetching

New ConnectX firmware introduces FW command interface revision 2,
which requires that for each QP, a chunk of send queue entries (the
"headroom") is kept marked as invalid, so that the HCA doesn't get
confused if it prefetches entries that haven't been posted yet.  Add
code to the driver to do this, and also update the user ABI so that
userspace can request that the prefetcher be turned off for userspace
QPs (we just leave the prefetcher on for all kernel QPs).

Unfortunately, marking send queue entries this way is confuses older
firmware, so we change the driver to allow only FW command interface
revisions 2.  This means that users will have to update their firmware
to work with the new driver, but the firmware is changing quickly and
the old firmware has lots of other bugs anyway, so this shouldn't be too
big a deal.

Based on a patch from Jack Morgenstein <jackm@dev.mellanox.co.il>.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agosh: oops_enter()/oops_exit() in die().
Paul Mundt [Mon, 18 Jun 2007 09:57:13 +0000 (18:57 +0900)]
sh: oops_enter()/oops_exit() in die().

As Russell helpfully pointed out on linux-arch:

http://marc.info/?l=linux-arch&m=118208089204630&w=2

We were missing the oops_enter/exit() in the sh die() implementation.
As we do support lockdep, it's beneficial to add these calls so lockdep
properly disables itself in the die() case.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
17 years agosh: Fix restartable syscall arg5 clobbering.
Kaz Kojima [Mon, 18 Jun 2007 01:08:20 +0000 (10:08 +0900)]
sh: Fix restartable syscall arg5 clobbering.

We use R0 as the 5th argument of syscall.  When the syscall restarts
after signal handling, we should restore the old value of R0.
The attached patch does it. Without this patch, I've experienced random
failures in the situation which signals are issued frequently.

Signed-off-by: Kaz Kojima <kkojima@rr.iij4u.or.jp>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
17 years agoLinux 2.6.22-rc5 v2.6.22-rc5
Linus Torvalds [Sun, 17 Jun 2007 02:09:12 +0000 (19:09 -0700)]
Linux 2.6.22-rc5

The manatees, they are dancing!

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoshm: fix the filename of hugetlb sysv shared memory
Eric W. Biederman [Sat, 16 Jun 2007 17:16:16 +0000 (10:16 -0700)]
shm: fix the filename of hugetlb sysv shared memory

Some user space tools need to identify SYSV shared memory when examining
/proc/<pid>/maps.  To do so they look for a block device with major zero, a
dentry named SYSV<sysv key>, and having the minor of the internal sysv
shared memory kernel mount.

To help these tools and to make it easier for people just browsing
/proc/<pid>/maps this patch modifies hugetlb sysv shared memory to use the
SYSV<key> dentry naming convention.

User space tools will still have to be aware that hugetlb sysv shared
memory lives on a different internal kernel mount and so has a different
block device minor number from the rest of sysv shared memory.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Albert Cahalan <acahalan@gmail.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agohugetlb: fix get_policy for stacked shared memory files
Adam Litke [Sat, 16 Jun 2007 17:16:15 +0000 (10:16 -0700)]
hugetlb: fix get_policy for stacked shared memory files

Here's another breakage as a result of shared memory stacked files :(

The NUMA policy for a VMA is determined by checking the following (in the
order given):

1) vma->vm_ops->get_policy() (if defined)
2) vma->vm_policy (if defined)
3) task->mempolicy (if defined)
4) Fall back to default_policy

By switching to stacked files for shared memory, get_policy() is now always
set to shm_get_policy which is a wrapper function.  This causes us to stop
at step 1, which yields NULL for hugetlb instead of task->mempolicy which
was the previous (and correct) result.

This patch modifies the shm_get_policy() wrapper to maintain steps 1-3 for
the wrapped vm_ops.

(akpm: the refcounting of mempolicies is busted and this patch does nothing to
improve it)

Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: William Irwin <bill.irwin@oracle.com>
Cc: dean gaudet <dean@arctic.org>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Andi Kleen <ak@suse.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoudf: fix possible leakage of blocks
Jan Kara [Sat, 16 Jun 2007 17:16:14 +0000 (10:16 -0700)]
udf: fix possible leakage of blocks

We have to take care that when we call udf_discard_prealloc() from
udf_clear_inode() we have to write inode ourselves afterwards (otherwise,
some changes might be lost leading to leakage of blocks, use of free blocks
or improperly aligned extents).

Also udf_discard_prealloc() does two different things - it removes
preallocated blocks and truncates the last extent to exactly match i_size.
We move the latter functionality to udf_truncate_tail_extent(), call
udf_discard_prealloc() when last reference to a file is dropped and call
udf_truncate_tail_extent() when inode is being removed from inode cache
(udf_clear_inode() call).

We cannot call udf_truncate_tail_extent() earlier as subsequent open+write
would find the last block of the file mapped and happily write to the end
of it, although the last extent says it's shorter.

[akpm@linux-foundation.org: Make checkpatch.pl happier]
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: Eric Sandeen <sandeen@sandeen.net>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoSLUB: minimum alignment fixes
Christoph Lameter [Sat, 16 Jun 2007 17:16:13 +0000 (10:16 -0700)]
SLUB: minimum alignment fixes

If ARCH_KMALLOC_MINALIGN is set to a value greater than 8 (SLUBs smallest
kmalloc cache) then SLUB may generate duplicate slabs in sysfs (yes again)
because the object size is padded to reach ARCH_KMALLOC_MINALIGN.  Thus the
size of the small slabs is all the same.

No arch sets ARCH_KMALLOC_MINALIGN larger than 8 though except mips which
for some reason wants a 128 byte alignment.

This patch increases the size of the smallest cache if
ARCH_KMALLOC_MINALIGN is greater than 8.  In that case more and more of the
smallest caches are disabled.

If we do that then the count of the active general caches that is displayed
on boot is not correct anymore since we may skip elements of the kmalloc
array.  So count them separately.

This approach was tested by Havard yesterday.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoRework ptep_set_access_flags and fix sun4c
Benjamin Herrenschmidt [Sat, 16 Jun 2007 17:16:12 +0000 (10:16 -0700)]
Rework ptep_set_access_flags and fix sun4c

Some changes done a while ago to avoid pounding on ptep_set_access_flags and
update_mmu_cache in some race situations break sun4c which requires
update_mmu_cache() to always be called on minor faults.

This patch reworks ptep_set_access_flags() semantics, implementations and
callers so that it's now responsible for returning whether an update is
necessary or not (basically whether the PTE actually changed).  This allow
fixing the sparc implementation to always return 1 on sun4c.

[akpm@linux-foundation.org: fixes, cleanups]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: David Miller <davem@davemloft.net>
Cc: Mark Fortescue <mark@mtfhpc.demon.co.uk>
Acked-by: William Lee Irwin III <wli@holomorphy.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agorandom: fix output buffer folding
Matt Mackall [Sat, 16 Jun 2007 17:16:11 +0000 (10:16 -0700)]
random: fix output buffer folding

(As reported by linux@horizon.com)

Folding is done to minimize the theoretical possibility of systematic
weakness in the particular bits of the SHA1 hash output.  The result of
this bug is that 16 out of 80 bits are un-folded.  Without a major new
vulnerability being found in SHA1, this is harmless, but still worth
fixing.

Signed-off-by: Matt Mackall <mpm@selenic.com>
Cc: <linux@horizon.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouml: kill x86_64 STACK_TOP_MAX
Jeff Dike [Sat, 16 Jun 2007 17:16:10 +0000 (10:16 -0700)]
uml: kill x86_64 STACK_TOP_MAX

The x86_64 a.out.h got a definition of STACK_TOP_MAX, which interferes with
the UML version.  So, just undef it like STACK_TOP.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouml: remove PAGE_SIZE from libc code
Jeff Dike [Sat, 16 Jun 2007 17:16:09 +0000 (10:16 -0700)]
uml: remove PAGE_SIZE from libc code

Distros seem to be removing PAGE_SIZE from asm/page.h.  So, the libc side of
UML should stop using it.

I replace it with UM_KERN_PAGE_SIZE, which is defined to be the same as
PAGE_SIZE on the kernel side of the house.  I could also use getpagesize(),
but it's more important that UML have the same value of PAGE_SIZE everywhere.
It's conceivable that it could be built with a larger PAGE_SIZE, and use of
getpagesize() would break that badly.

PAGE_MASK got the same treatment, as it is closely tied to PAGE_SIZE.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agospi doc updates
David Brownell [Sat, 16 Jun 2007 17:16:08 +0000 (10:16 -0700)]
spi doc updates

Update two points in the SPI interface documentation:

- Update description of the "chip stays selected after message ends"
  mode.  In some cases it's required for correctness; it isn't just a
  performance tweak.  (Yes: to use this mode on mult-device busses, another
  programming interface will be needed.  One draft has been circulated
  already.)

- Clarify spi_setup(), highlighting that callers must ensure that no
  requests are queued (can't change configuration except between I/Os), and
  that the device must be deselected when this returns (which is a key part
  of why it's called during device init).

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomd: fix bug in error handling during raid1 repair
Mike Accetta [Sat, 16 Jun 2007 17:16:07 +0000 (10:16 -0700)]
md: fix bug in error handling during raid1 repair

If raid1/repair (which reads all block and fixes any differences it finds)
hits a read error, it doesn't reset the bio for writing before writing
correct data back, so the read error isn't fixed, and the device probably
gets a zero-length write which it might complain about.

Signed-off-by: Neil Brown <neilb@suse.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomd: fix two raid10 bugs
NeilBrown [Sat, 16 Jun 2007 17:16:06 +0000 (10:16 -0700)]
md: fix two raid10 bugs

1/ When resyncing a degraded raid10 which has more than 2 copies of each block,
  garbage can get synced on top of good data.

2/ We round the wrong way in part of the device size calculation, which
  can cause confusion.

Signed-off-by: Neil Brown <neilb@suse.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofuse: ->fs_flags fixlet
Alexey Dobriyan [Sat, 16 Jun 2007 17:16:05 +0000 (10:16 -0700)]
fuse: ->fs_flags fixlet

fs/fuse/inode.c:658:3: error: Initializer entry defined twice
fs/fuse/inode.c:661:3:   also defined here

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoperfctr-watchdog: fix interchanged parameters to release_{evntsel,perfctr}_nmi
Björn Steinbrink [Sat, 16 Jun 2007 17:16:04 +0000 (10:16 -0700)]
perfctr-watchdog: fix interchanged parameters to release_{evntsel,perfctr}_nmi

Fix oops triggered during: echo 0 > /proc/sys/kernel/nmi_watchdog

The culprit seems to be 09198e68501a7e34737cd9264d266f42429abcdc:
[PATCH] i386: Clean up NMI watchdog code

In two places, the parameters to release_{evntsel,perfctr}_nmi
got interchanged during the cleanup.

Fix interchanged parameters to release_{evntsel,perfctr}_nmi.

Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de>
Cc: Stephane Eranian <eranian@hpl.hp.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Michal Piotrowski <michal.k.k.piotrowski@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoswsusp: Fix userland interface
Rafael J. Wysocki [Sat, 16 Jun 2007 17:16:03 +0000 (10:16 -0700)]
swsusp: Fix userland interface

Fix oops caused by 'cat /dev/snapshot', reported by Arkadiusz Miskiewicz,
and make it impossible to thaw tasks with the help of the swsusp userland
interface while there is a snapshot image ready to save.

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>