This patch adds support for the phyCORE-i.MX27 cpu module (aka pcm038).
It is as generic as possible in order to support any kind of baseboard.
Note: This CPU module implementation can't work without a baseboard
support. Baseboard support can be added by the PCM-970 (included in
this patch stack) or any custom variant.
Changed against the original Freescale code (and against clocklib for example):
- clock rate is always calculated whenever one ask for the current rate.
(means no "rate" member in the clock structure).
So switching the PLL base frequency will propagate immediately to all
other clocks that are depending on this frequency.
TODO:
- Check if the i.MX21 CPU can share the same code.
Daniel Mack [Sat, 5 Jul 2008 08:02:53 +0000 (10:02 +0200)]
MX3: Add basic support for LogicPD i.MX31 LiteKit
This patch adds basic support for i.MX31 LiteKit by LogicPD.
With printascii() in kernel/printk.c, it boots right into the
rootfs-panic.
Note: This is a modified version of Daniel's patch to fit into this patch
stack.
> On 09.06.2008, at 17:26, Russell King - ARM Linux wrote:
>
> > I would much prefer it if board specific includes were included by the
> > code which needs them rather than in asm/arch/hardware.h. With the
> > device model, drivers shouldn't need to include any board specific
> > includes - only the board specific C file should need it.
>
> The new version of this patch (#5102) has been uploaded to the patch
> tracker this morning.
TODO: It seems impossible to build a kernel for more than one CPU because the
timer do not follow the platform device rules. So it does only work if
timer 1 can be accessed on all CPUs at the same address.
This patch bases on the one from Daniel Mack. The most important change to
Daniel's patch is to be more generic. This gpio routine supports at least
the i.MX27 and i.MX31 processors.
Signed-off-by: Juergen Beisert <j.beisert@pengutronix.de> Acked-by: Daniel Mack <daniel@caiaq.de>
Changed against the original Freescale code (and against clocklib for example):
- clock rate is always calculated whenever one ask for the current rate
(means struct clk has no more a member called "rate"). So switching the PLL
base frequency will propagate immediately to all other clocks that are
depending on this frequency.
This patch removes the statically initialized tables from the i.MX serial
driver and makes the driver fully dependent on the information provided by
the platform_device.
We assumed a 16MHz reference frequency for the UART. While this
is true for i.MX1 most of the time it is not true for MX27/MX31.
Also, add handling for the ONEMS register which is present on
newer versions of the chip and pass a sane minimum baudrate to
uart_get_baud_rate().
* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
USB: adding comment for ipaq forcing number of ports
USB: fix Oops on loading ipaq module since 2.6.26
USB: add a pl2303 device id
USB: another option device id
USB: don't lose disconnections during suspend
USB: fix interrupt disabling for HCDs with shared interrupt handlers
USB: New device ID for ftdi_sio driver
sisusbvga: Fix oops on disconnect.
USB: mass storage: new id for US_SC_CYP_ATACB
USB: ohci - record data toggle after unlink
USB: ehci - fix timer regression
USB: fix cdc-acm resume()
OHCI: Fix problem if SM501 and another platform driver is selected
Stefan Becker [Tue, 1 Jul 2008 16:19:22 +0000 (19:19 +0300)]
USB: fix interrupt disabling for HCDs with shared interrupt handlers
USB: fix interrupt disabling for HCDs with shared interrupt handlers
As has been discussed several times on LKML, IRQF_SHARED | IRQF_DISABLED
doesn't work reliably, i.e. a shared interrupt handler CAN'T be certain to
be called with interrupts disabled. Most USB HCD handlers use IRQF_DISABLED
and therefore havoc can break out if they share their interrupt with a
handler that doesn't use it.
On my test machine the yenta_socket interrupt handler (no IRQF_DISABLED)
was registered before ehci_hcd and one uhci_hcd instance. Therefore all
usb_hcd_irq() invocations for ehci_hcd and for one uhci_hcd instance
happened with interrupts enabled. That led to random lockups as USB core
HCD functions that acquire the same spinlock could be called twice
from interrupt handlers.
This patch updates usb_hcd_irq() to always disable/restore interrupts.
usb_add_hcd() will silently remove any IRQF_DISABLED requested from HCD code.
Signed-off-by: Stefan Becker <stefan.becker@nokia.com> Cc: stable <stable@kernel.org> Acked-by: David Brownell <david-b@pacbell.net> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Jon K Hellan [Tue, 24 Jun 2008 09:43:13 +0000 (11:43 +0200)]
USB: New device ID for ftdi_sio driver
Here's a new device ID for the ftdio_sio driver.
The diff is with linus's tree as of this morning.
The device is the RigExpert Tiny USB Soundcard Transceiver Interface for ham
radio.
(I didn't actually test this. A fellow ham couldn't get the device to work, and
I suggested binding the device ID using sysfs - see
"http://jk.ufisa.uninett.no/usb/". However, he had had moved on to other things
by then. I guess adding the device ID to the kernel "on spec" won't hurt.
The relevant part of cat /proc/bus/usb/devices shows:
David Brownell [Sat, 14 Jun 2008 06:59:54 +0000 (23:59 -0700)]
USB: ohci - record data toggle after unlink
This patch fixes a problem with OHCI where canceling bulk or
interrupt URBs may lose track of the right data toggle. This
seems to be a longstanding bug, possibly dating back to the
Linux 2.4 kernel, which stayed hidden because
(a) about half the time the data toggle bit was correct;
(b) canceling such URBs is unusual; and
(c) the few drivers which cancel these URBs either
[1] do it only as part of shutting down, or
[2] have fault recovery logic, which recovers.
For those transfer types, the toggle is normally written back
into the ED when each TD is retired. But canceling bypasses
the mechanism used to retire TDs ... so on average, half the
time the toggle bit will be invalid after cancelation.
The fix is simple: the toggle state of any canceled TDs are
propagated back to the ED in the finish_unlinks function.
(Issue found by leonidv11@gmail.com ...)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Cc: Leonid <leonidv11@gmail.com> Cc: stable <stable@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
David Brownell [Sat, 14 Jun 2008 06:56:48 +0000 (23:56 -0700)]
USB: ehci - fix timer regression
This patch fixes a regression in the EHCI driver's TIMER_IO_WATCHDOG
behavior. The patch "USB: EHCI: add separate IAA watchdog timer" changed
how that timer is handled, so that short timeouts on the remaining
timer (unfortunately, overloaded) would never be used.
This takes a more direct approach, reorganizing the code slightly to
be explicit about only the I/O watchdog role now being overridable.
It also replaces a now-obsolete comment describing older timer behavior.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Cc: Alan Stern <stern@rowland.harvard.edu> Cc: Leonid <leonidv11@gmail.com> Cc: stable <stable@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Oliver Neukum [Tue, 17 Jun 2008 20:30:48 +0000 (22:30 +0200)]
USB: fix cdc-acm resume()
cdc-acm has
- a memory leak in resume()
- will fail to reactivate the read code path if this is needed.
his corrects it by deleting the useless relict code.
Signed-off-by: Oliver Neukum <oneukum@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Ben Dooks [Sun, 8 Jun 2008 16:20:11 +0000 (17:20 +0100)]
OHCI: Fix problem if SM501 and another platform driver is selected
If the SM501 and another platform driver, such as the SM501
then we end up defining PLATFORM_DRIVER twice. This patch
seperated the SM501 onto a seperate define of SM501_OHCI_DRIVER
so that it can be selected without overwriting the original
definition.
Signed-off-by: Ben Dooks <ben-linux@fluff.org> Acked-by: David Brownell <dbrownell@users.sourceforge.net> Cc: stable <stable@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
J. Bruce Fields [Thu, 3 Jul 2008 19:26:35 +0000 (15:26 -0400)]
svcrpc: fix handling of garbage args
To return garbage_args, the accept_stat must be 0, and we must have a
verifier. So we shouldn't be resetting the write pointer as we reject
the call.
Also, we must add the two placeholder words here regardless of success
of the unwrap, to ensure the output buffer is left in a consistent state
for svcauth_gss_release().
This fixes a BUG() in svcauth_gss.c:svcauth_gss_release().
Thanks to Aime Le Rouzic for bug report, debugging help, and testing.
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Tested-by: Aime Le Rouzic <aime.le-rouzic@bull.net> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Unfortunately this allows the task to migrate after setting up the softirq
and raising it. Since softirqs run a queue that is per-cpu we may raise the
softirq on the wrong CPU and this will keep the queued softirq task from
running.
To solve this issue, this patch disables preemption around the releasing
of the hrtimer lock and raising of the softirq.
Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Atsushi Nemoto [Tue, 24 Jun 2008 14:26:38 +0000 (23:26 +0900)]
[MIPS] cevt-txx9: Reset timer counter on initialization
The txx9_tmr_init() will not clear a timer counter register in a certain
case. The counter register is cleared on 1->0 transition of TCE bit if
CRE=1. So just clearing the TCE bit is not enough.
[MIPS] IP22: Fix crashes due to wrong L1_CACHE_BYTES
The introduction of a real dma cache invalidate makes it important
to have a correct cache line size, otherwise the kernel will gives
out two memory segment, which might share one cache line. The R4400
Indy/Indigo2 CPU modules are using a second level cache line size
of 128 bytes, so MIPS_L1_CACHE_SHIFT needs to be bumped up to 7 for
IP22.
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
It's possible that the crime interrupt handler is called without
pending interrupts (probably a hardware issue). To avoid irritating
"unexpected irq 71" messages, we now just ignore the spurious crime
interrupts.
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Do not overwrite nr_zones on !NUMA when initialising zlcache_ptr
The non-NUMA case of build_zonelist_cache() would initialize the
zlcache_ptr for both node_zonelists[] to NULL.
Which is problematic, since non-NUMA only has a single node_zonelists[]
entry, and trying to zero the non-existent second one just overwrote the
nr_zones field instead.
As kswapd uses this value to determine what reclaim work is necessary,
the result is that kswapd never reclaims. This causes processes to
stall frequently in low-memory situations as they always direct reclaim.
This patch initialises zlcache_ptr correctly.
Signed-off-by: Mel Gorman <mel@csn.ul.ie> Tested-by: Dan Williams <dan.j.williams@intel.com>
[ Simplified patch a bit ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The legacy protocol's open operation doesn't handle an append operation
(it is expected that the client take care of it). We were incorrectly
passing the extended protocol's flag through even in legacy mode. This
was reported in bugzilla report #10689. This patch fixes the problem
by disallowing extended protocol open modes from being passed in legacy
mode and implemented append functionality on the client side by adding
a seek after the open.
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
Merge branch 'i2c-fix' of git://aeryn.fluff.org.uk/bjdooks/linux
* 'i2c-fix' of git://aeryn.fluff.org.uk/bjdooks/linux:
I2C: S3C2410: Add MODULE_ALIAS() for s3c2440 device.
I2C: S3C2410: Fixup error codes returned rom a transfer.
I2C: S3C2410: Check ACK on byte transmission
Merge branch 'for-2.6.26' of git://git.kernel.dk/linux-2.6-block
* 'for-2.6.26' of git://git.kernel.dk/linux-2.6-block:
Properly notify block layer of sync writes
block: Fix the starving writes bug in the anticipatory IO scheduler
* git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/v4l-dvb:
V4L/DVB (8178): uvc: Fix compilation breakage for the other drivers, if uvc is selected
V4L/DVB (8145a): USB Video Class driver
* master.kernel.org:/home/rmk/linux-2.6-arm:
[ARM] 5131/1: Annotate platform_secondary_init with trace_hardirqs_off
[ARM] 5117/1: pxafb: fix __devinit/exit annotations
[ARM] Export dma_sync_sg_for_device()
[ARM] 5109/1: Mark rtc sa1100 driver as wakeup source before registering it
[ARM] 5116/1: pxafb: cleanup and fix order of failure handling
[ARM] 5115/1: pxafb: fix ifdef for command line option handling
ARM: OMAP: Correcting the gpmc prefetch control register address
ARM: OMAP: DMA: Don't mark channel active in omap_enable_channel_irq
Merge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
sched: fix divide error when trying to configure rt_period to zero
Merge branch 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6
* 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6:
i2c: Fix bad hint about irqs in i2c.h
i2c: Documentation: fix device matching description
Merge branch 'for-2.6.26' of git://neil.brown.name/md
* 'for-2.6.26' of git://neil.brown.name/md:
Fix error paths if md_probe fails.
Don't acknowlege that stripe-expand is complete until it really is.
Ensure interrupted recovery completed properly (v1 metadata plus bitmap)
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/mpc5200: Fix lite5200b suspend/resume
powerpc/legacy_serial: Bail if reg-offset/shift properties are present
powerpc/bootwrapper: update for initrd with simpleImage
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (55 commits)
net: fib_rules: fix error code for unsupported families
netdevice: Fix wrong string handle in kernel command line parsing
net: Tyop of sk_filter() comment
netlink: Unneeded local variable
net-sched: fix filter destruction in atm/hfsc qdisc destruction
net-sched: change tcf_destroy_chain() to clear start of filter list
ipv4: fix sysctl documentation of time related values
mac80211: don't accept WEP keys other than WEP40 and WEP104
hostap: fix sparse warnings
hostap: don't report useless WDS frames by default
textsearch: fix Boyer-Moore text search bug
netfilter: nf_conntrack_tcp: fixing to check the lower bound of valid ACK
ipv6 route: Convert rt6_device_match() to use RT6_LOOKUP_F_xxx flags.
netlabel: Fix a problem when dumping the default IPv6 static labels
net/inet_lro: remove setting skb->ip_summed when not LRO-able
inet fragments: fix race between inet_frag_find and inet_frag_secret_rebuild
CONNECTOR: add a proc entry to list connectors
netlink: Fix some doc comments in net/netlink/attr.c
tcp: /proc/net/tcp rto,ato values not scaled properly (v2)
include/linux/netdevice.h: don't export MAX_HEADER to userspace
...
When scheduled swaps occur, we need to blit between front & back
buffers. If the buffers are tiled, we need to set the appropriate
XY_SRC_COPY tile bit, but only on 965 chips, since it will cause
corruption on pre-965 (e.g. 945).
Bug reported by and fix tested by Tomas Janousek <tomi@nomi.cz>.
| linux/drivers/input/ff-core.c: In function 'input_ff_upload':
| linux/drivers/input/ff-core.c:172: error: dereferencing pointer to incomplete type
| linux/drivers/input/ff-core.c: In function 'erase_effect':
| linux/drivers/input/ff-core.c:197: error: dereferencing pointer to incomplete type
| linux/drivers/input/ff-core.c:204: error: dereferencing pointer to incomplete type
| make[4]: *** [drivers/input/ff-core.o] Error 1
As the incomplete type is `struct task_struct', including <linux/sched.h> fixes
it.
Alex Chiang [Wed, 2 Jul 2008 02:02:23 +0000 (20:02 -0600)]
PCI: acpiphp: cleanup notify handler on all root bridges
During the development of the physical PCI slot patch series, Gary Hade
kept on reporting strange oopses due to interactions between pci_slot
and acpiphp.
http://lkml.org/lkml/2007/11/28/319
find_root_bridges() unconditionally installs
handle_hotplug_event_bridge() as an ACPI_SYSTEM_NOTIFY handler for all
root bridges.
However, during module cleanup, remove_bridge() will only remove the
notify handler iff the root bridge had a hot-pluggable slot directly
underneath. That is:
root bridge -> hotplug slot
But, if the topology looks like either of the following:
Then we currently do not remove the notify handler from that root
bridge.
This can cause a kernel oops if we modprobe acpiphp later and it gets
loaded somewhere else in memory. If the root bridge then receives a
hotplug event, it will then attempt to call a stale, non-existent notify
handler and we blow up.
Much thanks goes to Gary Hade for his persistent debugging efforts.
Signed-off-by: Alex Chiang <achiang@hp.com> Signed-off-by: Gary Hade <garyhade@us.ibm.com> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
For Broadcom 5706, 5708, 5709 rev. A nics, any read beyond the
VPD end tag will hang the device. This problem was initially
observed when a vpd entry was created in sysfs
('/sys/bus/pci/devices/<id>/vpd'). A read to this sysfs entry
will dump 32k of data. Reading a full 32k will cause an access
beyond the VPD end tag causing the device to hang. Once the device
is hung, the bnx2 driver will not be able to reset the device.
We believe that it is legal to read beyond the end tag and
therefore the solution is to limit the read/write length.
A majority of this patch is from Matthew Wilcox who gave code for
reworking the PCI vpd size information. A PCI quirk added for the
Broadcom NIC's to limit the read/write's.
Signed-off-by: Benjamin Li <benli@broadcom.com> Signed-off-by: Matthew Wilcox <willy@linux.intel.com> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Milan Broz [Wed, 2 Jul 2008 08:34:28 +0000 (09:34 +0100)]
dm crypt: use cond_resched
Add cond_resched() to prevent monopolising CPU when processing large bios.
dm-crypt processes encryption of bios in sector units. If the bio request
is big it can spend a long time in the encryption call.
Signed-off-by: Milan Broz <mbroz@redhat.com> Tested-by: Yan Li <elliot.li.tech@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Wang Chen [Wed, 2 Jul 2008 02:57:19 +0000 (19:57 -0700)]
netdevice: Fix wrong string handle in kernel command line parsing
v1->v2: Use strlcpy() to ensure s[i].name be null-termination.
1. In netdev_boot_setup_add(), a long name will leak.
ex. : dev=21,0x1234,0x1234,0x2345,eth123456789verylongname.........
2. In netdev_boot_setup_check(), mismatch will happen if s[i].name
is a substring of dev->name.
ex. : dev=...eth1 dev=...eth11
[ With feedback from Ben Hutchings. ]
Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
ipv4: fix sysctl documentation of time related values
These sysctl values are time related and all use the same routine
(proc_dointvec_jiffies) that internally converts from seconds to jiffies.
The code is fine, the documentation is just wrong.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Tim Yamin [Tue, 17 Jun 2008 08:33:14 +0000 (09:33 +0100)]
powerpc/mpc5200: Fix lite5200b suspend/resume
Suspend/resume ("echo mem > /sys/power/state") does not work with
vanilla kernels -- the system does not suspend correctly and just
hangs. This patch fixes this so suspend/resume works:
1) of_iomap does not map the whole 0xC000 of the MPC5200 immr so
saving registers does not work.
2) PCI registers need to be saved and restored.
Signed-off-by: Tim Yamin <plasm@roo.me.uk> Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
John Linn [Tue, 1 Jul 2008 17:52:41 +0000 (10:52 -0700)]
powerpc/legacy_serial: Bail if reg-offset/shift properties are present
The legacy serial driver does not work with an 8250 type UART that is
described in the device tree with the reg-offset and reg-shift
properties. This change makes legacy_serial ignore these devices.
Signed-off-by: John Linn <john.linn@xilinx.com> Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Wolfram Sang [Tue, 1 Jul 2008 20:38:18 +0000 (22:38 +0200)]
i2c: Fix bad hint about irqs in i2c.h
i2c.h mentions -1 as a not-issued irq. This false hint was taken by
of_i2c and caused crashes. Don't give any advice as 'no irq' is not
consistent across all architectures yet and it is not needed internally
by the i2c-core.
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de> Signed-off-by: Jean Delvare <khali@linux-fr.org>
The matching process described for new style clients in
Documentation/i2c/writing-clients is classed as out-of-date
as it requires the presence of an .id_table entry in the
driver's i2c_driver entry.
Signed-off-by: Ben Dooks <ben-linux@fluff.org> Signed-off-by: Jean Delvare <khali@linux-fr.org>
Ben Hutchings [Tue, 1 Jul 2008 16:18:17 +0000 (17:18 +0100)]
PCI: Restrict VPD read permission to root
Some PCI devices will lock up if we attempt to read from VPD addresses
beyond some device-dependent limit. Until we can identify these
devices and adjust the file size accordingly, only let root read VPD
through sysfs to prevent a DoS by normal users.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Ben Dooks [Tue, 1 Jul 2008 10:59:41 +0000 (11:59 +0100)]
I2C: S3C2410: Check ACK on byte transmission
We should check for the reception of an ACK after transmitting each
data byte. The address send has been correctly checking this, but the
data write byte state should have also been checking for these failures.
As part of the same fix, we remove the ACK checking from the receive
path where it should not have been checking for an ACK which our hardware
was sending.
Gautham R Shenoy [Fri, 27 Jun 2008 04:47:38 +0000 (10:17 +0530)]
rcu: fix hotplug vs rcu race
Dhaval Giani reported this warning during cpu hotplug stress-tests:
| On running kernel compiles in parallel with cpu hotplug:
|
| WARNING: at arch/x86/kernel/smp.c:118
| native_smp_send_reschedule+0x21/0x36()
| Modules linked in:
| Pid: 27483, comm: cc1 Not tainted 2.6.26-rc7 #1
| [...]
| [<c0110355>] native_smp_send_reschedule+0x21/0x36
| [<c014fe8f>] force_quiescent_state+0x47/0x57
| [<c014fef0>] call_rcu+0x51/0x6d
| [<c01713b3>] __fput+0x130/0x158
| [<c0171231>] fput+0x17/0x19
| [<c016fd99>] filp_close+0x4d/0x57
| [<c016fdff>] sys_close+0x5c/0x97
IMHO the warning is a spurious one.
cpu_online_map is updated by the _cpu_down() using stop_machine_run().
Since force_quiescent_state is invoked from irqs disabled section,
stop_machine_run() won't be executing while a cpu is executing
force_quiescent_state(). Hence the cpu_online_map is stable while we're
in the irq disabled section.
However, a cpu might have been offlined _just_ before we disabled irqs
while entering force_quiescent_state(). And rcu subsystem might not yet
have handled the CPU_DEAD notification, leading to the offlined cpu's
bit being set in the rcp->cpumask.
Hence cpumask = (rcp->cpumask & cpu_online_map) to prevent sending
smp_reschedule() to an offlined CPU.
Here's the timeline:
CPU_A CPU_B
--------------------------------------------------------------
cpu_down(): .
. .
. .
stop_machine(): /* disables preemption, .
* and irqs */ .
. .
. .
take_cpu_down(); .
. .
. .
. .
cpu_disable(); /*this removes cpu .
*from cpu_online_map .
*/ .
. .
. .
restart_machine(); /* enables irqs */ .
------WINDOW DURING WHICH rcp->cpumask is stale ---------------
. call_rcu();
. /* disables irqs here */
. .force_quiescent_state();
.CPU_DEAD: .for_each_cpu(rcp->cpumask)
. . smp_send_reschedule();
. .
. . WARN_ON() for offlined CPU!
.
.
.
rcu_cpu_notify:
.
-------- WINDOW ENDS ------------------------------------------
rcu_offline_cpu() /* Which calls cpu_quiet()
* which removes
* cpu from rcp->cpumask.
*/
If a new batch was started just before calling stop_machine_run(), the
"tobe-offlined" cpu is still present in rcp-cpumask.
During a cpu-offline, from take_cpu_down(), we queue an rt-prio idle
task as the next task to be picked by the scheduler. We also call
cpu_disable() which will disable any further interrupts and remove the
cpu's bit from the cpu_online_map.
Once the stop_machine_run() successfully calls take_cpu_down(), it calls
schedule(). That's the last time a schedule is called on the offlined
cpu, and hence the last time when rdp->passed_quiesc will be set to 1
through rcu_qsctr_inc().
But the cpu_quiet() will be on this cpu will be called only when the
next RCU_SOFTIRQ occurs on this CPU. So at this time, the offlined CPU
is still set in rcp->cpumask.
Now coming back to the idle_task which truely offlines the CPU, it does
check for a pending RCU and raises the softirq, since it will find
rdp->passed_quiesc to be 0 in this case. However, since the cpu is
offline I am not sure if the softirq will trigger on the CPU.
Even if it doesn't the rcu_offline_cpu() will find that rcp->completed
is not the same as rcp->cur, which means that our cpu could be holding
up the grace period progression. Hence we call cpu_quiet() and move
ahead.
But because of the window explained in the timeline, we could still have
a call_rcu() before the RCU subsystem executes it's CPU_DEAD
notification, and we send smp_send_reschedule() to offlined cpu while
trying to force the quiescent states. The appended patch adds comments
and prevents checking for offlined cpu everytime.
cpu_online_map is updated by the _cpu_down() using stop_machine_run().
Since force_quiescent_state is invoked from irqs disabled section,
stop_machine_run() won't be executing while a cpu is executing
force_quiescent_state(). Hence the cpu_online_map is stable while we're
in the irq disabled section.
fsync_buffers_list() and sync_dirty_buffer() both issue async writes and
then immediately wait on them. Conceptually, that makes them sync writes
and we should treat them as such so that the IO schedulers can handle
them appropriately.
This patch fixes a write starvation issue that Lin Ming reported, where
xx is stuck for more than 2 minutes because of a large number of
synchronous IO in the system:
Lin Ming confirms that this patch fixes the issue. I've run tests with
it for the past week and no ill effects have been observed, so I'm
proposing it for inclusion into 2.6.26.
Divyesh Shah [Mon, 16 Jun 2008 16:37:08 +0000 (18:37 +0200)]
block: Fix the starving writes bug in the anticipatory IO scheduler
AS scheduler alternates between issuing read and write batches. It does
the batch switch only after all requests from the previous batch are
completed.
When switching to a write batch, if there is an on-going read request,
it waits for its completion and indicates its intention of switching by
setting ad->changed_batch and the new direction but does not update the
batch_expire_time for the new write batch which it does in the case of
no previous pending requests.
On completion of the read request, it sees that we were waiting for the
switch and schedules work for kblockd right away and resets the
ad->changed_data flag.
Now when kblockd enters dispatch_request where it is expected to pick
up a write request, it in turn ends the write batch because the
batch_expire_timer was not updated and shows the expire timestamp for
the previous batch.
This results in the write starvation for all the cases where there is
the intention for switching to a write batch, but there is a previous
in-flight read request and the batch gets reverted to a read_batch
right away.
This also holds true in the reverse case (switching from a write batch
to a read batch with an in-flight write request).
I've checked that this bug exists on 2.6.11, 2.6.18, 2.6.24 and
linux-2.6-block git HEAD. I've tested the fix on x86 platforms with
SCSI drives where the driver asks for the next request while a current
request is in-flight.
This patch is based off linux-2.6-block git HEAD.
Bug reproduction:
A simple scenario which reproduces this bug is:
- dd if=/dev/hda3 of=/dev/null &
- lilo
The lilo takes forever to complete.
This can also be reproduced fairly easily with the earlier dd and
another test
program doing msync().
The example test program below should print out a message after every
iteration
but it simply hangs forever. With this bugfix it makes forward progress.
====
Example test program using msync() (thanks to suleiman AT google DOT
com)
I'm checking as soon as possible for the period not being zero since, if
it is, going ahead is useless. This way we also save a mutex_lock() and
a read_lock() wrt doing it inside tg_set_bandwidth() or
__rt_schedulable().
Tony Luck [Mon, 30 Jun 2008 22:03:14 +0000 (15:03 -0700)]
[IA64] Bugfix for system with 32 cpus
On a system where there are no hot pluggable cpus "additional_cpus"
is still set to -1 at the point where we call per_cpu_scan_finalize().
If we didn't find an SRAT table and so pick the default "32" for the
number of cpus, when we get to:
high_cpu = min(high_cpu + reserve_cpus, NR_CPUS);
we will end up initializing for just 31 cpus ... and so we will
die horribly when bringing up cpu#32.
Laurent Pinchart [Mon, 30 Jun 2008 18:04:50 +0000 (15:04 -0300)]
V4L/DVB (8145a): USB Video Class driver
This driver supports video input devices compliant with the USB Video Class
specification. This means lots of currently manufactured webcams, and probably
most of the future ones.
projects / linux-2.6-omap-h63xx.git / log