- add result register #defines
- rename ds_dump_status to ds_print_msg
- rename ds_recv_status to ds_dump_status
- ds_dump_status prints the requested status and no longer reads the
  status, this is because the second status read can return different
  data for example the result register
- the result register will be printed, though limited to detecting a
  new device, detecting other values such as a short would require
  additional reporting methods
- ST_EPOF was moved to ds_wait_status to clear the error condition
  sooner
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Simplify and fix ds_touch_bit.  If a device is attached in the middle of a
bus search the status register will return more than the default 16 bytes.
 The additional bytes indicate that it has detected a new device.  The way
ds_wait_status is coded, if it doesn't read 16 status bytes it returns an
error value.  ds_touch_bit then will detect that error and return an
error.  In that case it doesn't read the input buffer and returns
uninitialized data.  It doesn't stop there.  The next transaction will not
expect the extra byte in the input buffer and the short read will cause an
error and clear out both the old byte and new data in the input buffer.

Just ignore the value of ds_wait_status.  It is still required to wait
until ds2490 is again idle and there is data to read when ds_recv_data is
called.  This also removes the while loop.  None of the other commands
wait and verify that the issued command is in the status register.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Don't export read and write bit operations, they didn't work, they weren't
used, and they can't be made to work.  The one wire low level bit
operations expect to set high or low levels, the ds2490 hardware only
supports complete read or write time slots, better to just comment them
out.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

ds_write_bit doesn't read the input buffer, so add COMM_ICP and a comment
that it will no longer generate a read back data byte.  If there is an
extra data byte later on then it will cause an error and discard what data
was there.  Corrected operator ordering for ds_send_control.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add strong pullup support for ds2490 driver, also drop mdelay(750), which
busy waits, usage in favour of msleep for long delays.  Now with msleep
only being called when the strong pullup is active, one wire bus
operations are only taking minimal system overhead.

The new set_pullup will only enable the strong pullup when requested,
which is expected to be the only write operation that will benefit from a
strong pullup.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Corrected print message, it was writing not reading, this also prints the
endpoint used for the write instead of hardcoding it.  Failed to write
1-wire data to ep0x%x: err=%d.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Like the previous w1_io.c reset coments and msleep patch, I don't have the
hardware to verify the change, but I think it is safe.  It also helps to
see a comment like this in the code.  "We'll wait a bit longer just to be
sure." If they are going to calculate delaying 324.9us, but actually delay
500us, why not just give up the CPU and sleep?  This is designed for a
battery powered ARM system, avoiding busywaiting has to be good for
battery life.

I sent a request for testers March 7, 2008 to the Linux kernel mailing
list and two developers who have patches for ds1wm.c, but I didn't get
any respons.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

w1_reset_bus, added some comments about the timing and switched to msleep
for the later delay.  I don't have the hardware to test the sleep after
reset change.  The one wire doesn't have a timing requirement between
commands so it is fine.  I do have the USB hardware and it would be in big
trouble with 10ms interrupt transfers to find that the reset completed.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

s/printk/dev_dbg/
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Follow the example of other devices (like the joystick device).  Pick the
first available id for each detected device.  Currently for USB devices,
suspending and resuming would cause the number to increment.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

sl->master->mutex and dev->mutex refer to the same mutex variable, but be
consistent and use the same set of pointers for the lock and unlock calls.
 It is less confusing (and one less pointer dereference this way).
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Removed the w1_family structure member variable need_exit.  It was only
being set and never used.  Even if it were to be used it is a polling type
operation.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fixed data reading bug by replacing binary attribute with device one.

Switching the sysfs read from bin_attribute to device_attribute.  The data
is far under PAGE_SIZE so the binary interface isn't required.  As the
device_attribute interface will make one call to w1_therm_read per file
open and buffer, the result is, the following problems go away.

buffer overflow:
	Execute a short read on w1_slave and w1_therm_read_bin would still
	return the full string size worth of data clobbering the user space
	buffer when it returned.  Switching to device_attribute avoids the
	buffer overflow problems.  With the snprintf formatted output dealing
	with short reads without doing a conversion per read would have
	been difficult.
bad behavior:
	`cat w1_slave` would cause two temperature conversions to take place.
	Previously the code assumed W1_SLAVE_DATA_SIZE would be returned with
	each read.  It would not return 0 unless the offset was less
	than W1_SLAVE_DATA_SIZE.  The result was the first read did a
	temperature conversion, filled the buffer and returned, the
	offset in the second read would be less than
	W1_SLAVE_DATA_SIZE and also fill the buffer and return, the
	third read would finnally have a big enough offset to return 0
	and cause cat to stop.  Now w1_therm_read will be called at
	most once per open.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fix bug reading the id sysfs file.  If less than the full 8 bytes were
read, the next read would start at the first byte instead of continuing.
It needed the offset added to memcpy, or the better solution was to
replace it with the device attribute instead of bin attribute.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Document w1_master_add, w1_master_remove, search_count, and pullup.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Added a new module parameter search_count which allows overriding the
default search count.  -1 continual, 0 disabled, N that many times.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Simplified the logic in w1_slave_found by using the new
w1_attach_slave_device function to find a slave and mark it as active or
add the device if the crc checks.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

sysfs entries were added to manually add and remove slave devices.  This
is useful if the automatic bus searching is disabled, and the device ids
are already known.

[akpm@linux-foundation.org: fix printk types]
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Added strong pullup to thermal sensor driver and general documentation on
the sensor.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add a strong pullup option to the w1 system.  This supplies extra power
for parasite powered devices.  There is a w1_master_pullup sysfs entry and
enable_pullup module parameter to enable or disable the strong pullup.

The one wire bus requires at a minimum one wire and ground.  The common
wire is used for sending and receiving data as well as supplying power to
devices that are parasite powered of which temperature sensors can be one
example.  The bus must be idle and left high while a temperature
conversion is in progress, in addition the normal pullup resister on
larger networks or even higher temperatures might not supply enough power.
 The pullup resister can't provide too much pullup current, because
devices need to pull the bus down to write a value.  This enables the
strong pullup for supported hardware, which can supply more current when
requested.  Unsupported hardware will just delay with the bus high.

The hardware USB 2490 one wire bus master has a bit on some commands which
will enable the strong pullup as soon as the command finishes executing.
To use strong pullup, call the new w1_next_pullup function to register the
duration.  The next write command will call set_pullup before sending the
data, and reset the duration to zero once it returns.

Switched from simple_strtol to strict_strtol.
Signed-off-by: NDavid Fries <david@fries.net>
Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The w1_process thread's sleeping and termination has been modified.
msleep_interruptible was replaced by schedule_timeout and schedule to
allow for kthread_stop and wake_up_process to interrupt the sleep and the
unbounded sleeping when a bus search is disabled.  The W1_MASTER_NEED_EXIT
and flags variable were removed as they were redundant with
kthread_should_stop and kthread_stop.  If w1_process is sleeping,
requesting a search will immediately wake it up rather than waiting for
the end of msleep_interruptible previously.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Move the creation of the w1_process thread to after the device has been
initialized.  This way w1_process doesn't have to check to see if it has
been initialized and the bus search can proceed without sleeping.  That
also eliminates two checks in the w1_process loop.  The sleep now happens
at the end of the loop not the beginning.

Also added a comment for why the atomic_set was 2.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Early abort if the master driver or the hardware goes away in the middle
of a bus search operation.  The alternative is to spam the print buffer up
to 64*64 times with read errors in the case of USB.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

w1_control_thread was removed which would wake up every second and process
newly registered family codes and complete some final cleanup for a
removed master.  Those routines were moved to the threads that were
previously requesting those operations.  A new function
w1_reconnect_slaves takes care of reconnecting existing slave devices when
a new family code is registered or removed.  The removal case was missing
and would cause a deadlock waiting for the family code reference count to
decrease, which will now happen.  A problem with registering a family code
was fixed.  A slave device would be unattached if it wasn't yet claimed,
then attached at the end of the list, two unclaimed slaves would cause an
infinite loop.

The struct w1_bus_master.search now takes a pointer to the struct
w1_master device to avoid searching for it, which would have caused a
lock ordering deadlock with the removal of w1_control_thread.
Signed-off-by: NDavid Fries <david@fries.net>
Signed-off-by: NEvgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

tpm_register_hardware() leaks devname on an error path.

Addresses http://bugzilla.kernel.org/show_bug.cgi?id=11425Reported-by: NDaniel Marjamki <danielm77@spray.se>
Cc: Debora Velarde <debora@linux.vnet.ibm.com>
Cc: Rajiv Andrade <srajiv@linux.vnet.ibm.com>
Cc: Marcel Selhorst <tpm@selhorst.net>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Cc: Eric Paris <eparis@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Patch tpm-correct-tpm-timeouts-to-jiffies-conversion reveals a bug in the
Broadcom BCM0102 TPM chipset used in the Dell Latitude D820 - although
most of the timeouts are returned in usecs as per the spec, one is
apparently returned in msecs, which results in a too-small value leading
to a timeout when the code treats it as usecs.  To prevent a regression,
we check for the known too-short value and adjust it to a value that makes
things work.
Signed-off-by: NValdis Kletnieks <valdis.kletnieks@vt.edu>
Cc: Marcin Obara <marcin_obara@users.sourceforge.net>
Cc: Marcel Selhorst <tpm@selhorst.net>
Cc: Kylene Jo Hall <kjhall@us.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patch fixes timeouts conversion to jiffies, by replacing
msecs_to_jiffies() calls with usecs_to_jiffies().  According to TCG TPM
Specification Version 1.2 Revision 103 (pages 166, 167) TPM timeouts and
durations are returned in microseconds (usec) not in miliseconds (msec).

This fixes a long hang while loading TPM driver, if TPM chip starts in
"Idle" state instead of "Ready" state.  Without this patch - 'modprobe'
may hang for 30 seconds or more.
Signed-off-by: NMarcin Obara <marcin_obara@users.sourceforge.net>
Cc: Marcel Selhorst <tpm@selhorst.net>
Cc: Kylene Jo Hall <kjhall@us.ibm.com>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Remove CVS keywords that weren't updated for a long time from comments.
Signed-off-by: NAdrian Bunk <bunk@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Cannot assume writes will fully complete, so this conversion goes the easy
way and always brings the page uptodate before the write.

[dhowells@redhat.com: style tweaks]
Signed-off-by: NNick Piggin <npiggin@suse.de>
Acked-by: NDavid Howells <dhowells@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Make the Thermal messages (temperature got past Tmid) be displayed only
once because:

1) it's the BIOS job to configure and handle the memory throttling
2) if the BIOS is broken or is aware about the condition, flooding the
   system logs won't help anything.
3) According to the specification update for Intel 5000 MCHs, all the
   revisions of this MCH have problems on the thermal sensors, making
   not automatic (a.k.a. intelligent thermal throttling) impossible.
Signed-off-by: NAristeu Rozanski <aris@redhat.com>
Signed-off-by: NDoug Thompson <dougthompson@xmission.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Update the i5000_edac messages, making everything pass through the EDAC
(so the log controls will work) and being more specific about the errors.
Also, it makes the miscellaneous errors optional and disabled by default.

As I didn't found anywhere information about M23ERR-M26ERR
(FERR_NF_THERMAL) on FERR_NF_FBD, I'm removing them.
Signed-off-by: NAristeu Rozanski <aris@redhat.com>
Signed-off-by: NDoug Thompson <dougthompson@xmission.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This adds support for the dual-core MPC8572 processor.  We have
to support making SPR changes on each core.  Also, since we can
have multiple memory controllers sharing an interrupt, flag the
interrupts with IRQF_SHARED.
Signed-off-by: NAndrew Kilkenny <akilkenny@xes-inc.com>
Signed-off-by: NNate Case <ncase@xes-inc.com>
Acked-by: NDave Jiang <djiang@mvista.com>
Signed-off-by: NDoug Thompson <dougthompson@xmission.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fix 443BX/GX MCH suppport in a EDAC.

It makes i82443bxgx_edac coexist with intel_agp using the same approach as
several other EDAC drivers.

Tested on Intel's L443GX with redhat's 2.6.18 with whole EDAC subsystem
backported a while ago.

[root@host ~]# dmesg|grep -iE '(AGP|EDAC)'
Linux agpgart interface v0.101 (c) Dave Jones
agpgart: Detected an Intel 440GX Chipset.
agpgart: AGP aperture is 64M @ 0xf8000000
EDAC MC: Ver: 2.1.0 Jun 27 2008
EDAC MC0: Giving out device to 'i82443bxgx_edac' 'I82443BXGX': DEV 0000:00:00.0
EDAC PCI0: Giving out device to module 'i82443bxgx_edac' controller 'EDAC PCI controller': DEV '0000:00:00.0' (POLLED)
Signed-off-by: NVladislav Bogdanov <slava@nsys.by>
Cc: Doug Thompson <norsk5@yahoo.com>
Cc: Dave Airlie <airlied@linux.ie>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

task_child_reaper() has no callers anymore, kill it.
Signed-off-by: NOleg Nesterov <oleg@tv-sign.ru>
Acked-by: NSerge Hallyn <serue@us.ibm.com>
Acked-by: NPavel Emelyanov <xemul@openvz.org>
Acked-by: NSukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

de_thread() checks if the old leader was the ->child_reaper, this is not
possible any longer.  With the previous patch ->group_leader itself will
change ->child_reaper on exit.

Henceforth find_new_reaper() is the only function (apart from
initialization) which plays with ->child_reaper.
Signed-off-by: NOleg Nesterov <oleg@tv-sign.ru>
Acked-by: NSerge Hallyn <serue@us.ibm.com>
Acked-by: NPavel Emelyanov <xemul@openvz.org>
Acked-by: NSukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

name and nlen parameters passed to ->strategy hook are unused, remove
them.  In general ->strategy hook should know what it's doing, and don't
do something tricky for which, say, pointer to original userspace array
may be needed (name).
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net> [ networking bits ]
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Matt Mackall <mpm@selenic.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Move it into sysrq.c, along with the rest of the sysrq implementation.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

We currently follow blindly what the partition table lies about the
disk, and let the kernel create block devices which can not be accessed.
Trying to identify the device leads to kernel logs full of:
  sdb: rw=0, want=73392, limit=28800
  attempt to access beyond end of device

Here is an example of a broken partition table, where sda2 starts
behind the end of the disk, and sdb3 is larger than the entire disk:
  Disk /dev/sdb: 14 MB, 14745600 bytes
  1 heads, 29 sectors/track, 993 cylinders, total 28800 sectors
     Device Boot      Start         End      Blocks   Id  System
  /dev/sdb1              29        7800        3886   83  Linux
  /dev/sdb2           37801       45601        3900+  83  Linux
  /dev/sdb3           15602       73402       28900+  83  Linux
  /dev/sdb4           23403       28796        2697   83  Linux

The kernel creates these completely invalid devices, which can not be
accessed, or may lead to other unpredictable failures:
  grep . /sys/class/block/sdb*/{start,size}
  /sys/class/block/sdb/size:28800
  /sys/class/block/sdb1/start:29
  /sys/class/block/sdb1/size:7772
  /sys/class/block/sdb2/start:37801
  /sys/class/block/sdb2/size:7801
  /sys/class/block/sdb3/start:15602
  /sys/class/block/sdb3/size:57801
  /sys/class/block/sdb4/start:23403
  /sys/class/block/sdb4/size:5394

With this patch, we ignore partitions which start behind the end of the disk,
and limit partitions to the end of the disk if they pretend to be larger:
  grep . /sys/class/block/sdb*/{start,size}
  /sys/class/block/sdb/size:28800
  /sys/class/block/sdb1/start:29
  /sys/class/block/sdb1/size:7772
  /sys/class/block/sdb3/start:15602
  /sys/class/block/sdb3/size:13198
  /sys/class/block/sdb4/start:23403
  /sys/class/block/sdb4/size:5394

These warnings are printed to the kernel log:
  sdb: p2 ignored, start 37801 is behind the end of the disk
  sdb: p3 size 57801 limited to end of disk
Signed-off-by: NKay Sievers <kay.sievers@vrfy.org>
Cc: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

I missed this when I did the arm26 removal.
Reported-by: NRobert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: NAdrian Bunk <bunk@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>