When a new device is added below a hotplug bridge, the bridge's secondary
bus speed and the device's bus speed must match. The shpchp driver
previously checked the bridge's *primary* bus speed, not the secondary bus
speed.
This caused hot-add errors like:
shpchp 0000:00:03.0: Speed of bus ff and adapter 0 mismatch
Check the secondary bus speed instead.
[bhelgaas: changelog] Link: https://bugzilla.kernel.org/show_bug.cgi?id=75251 Fixes: 3749c51ac6c1 ("PCI: Make current and maximum bus speeds part of the PCI core") Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Patch 01f8fa4f01d "genirq: Allow forcing cpu affinity of interrupts" added
an irq_force_affinity() function, and 30ccf03b4a6 "clocksource: Exynos_mct:
Use irq_force_affinity() in cpu bringup" subsequently uses it. However, the
driver can be used with CONFIG_SMP disabled, but the function declaration
is only available for CONFIG_SMP, leading to this build error:
drivers/clocksource/exynos_mct.c:431:3: error: implicit declaration of function 'irq_force_affinity' [-Werror=implicit-function-declaration]
irq_force_affinity(mct_irqs[MCT_L0_IRQ + cpu], cpumask_of(cpu));
This patch introduces a dummy helper function for the non-SMP case
that always returns success, to get rid of the build error.
Since the patches causing the problem are marked for stable backports,
this one should be as well.
Signed-off-by: Arnd Bergmann <arnd@arndb.de> Cc: Krzysztof Kozlowski <k.kozlowski@samsung.com> Acked-by: Kukjin Kim <kgene.kim@samsung.com> Link: http://lkml.kernel.org/r/5619084.0zmrrIUZLV@wuerfel Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
b3b42ac2cbae x86-64, modify_ldt: Ban 16-bit segments on 64-bit kernels
disabled 16-bit segments on 64-bit kernels due to an information
leak. However, it does seem that people are genuinely using Wine to
run old 16-bit Windows programs on Linux.
A proper fix for this ("espfix64") is coming in the upcoming merge
window, but as a temporary fix, create a sysctl to allow the
administrator to re-enable support for 16-bit segments.
It adds a "/proc/sys/abi/ldt16" sysctl that defaults to zero (off). If
you hit this issue and care about your old Windows program more than
you care about a kernel stack address information leak, you can do
echo 1 > /proc/sys/abi/ldt16
as root (add it to your startup scripts), and you should be ok.
The sysctl table is only added if you have COMPAT support enabled on
x86-64, but I assume anybody who runs old windows binaries very much
does that ;)
Specify the maximum stack size for arches where the stack grows upward
(parisc and metag) in asm/processor.h rather than hard coding in
fs/exec.c so that metag can specify a smaller value of 256MB rather than
1GB.
This fixes a BUG on metag if the RLIMIT_STACK hard limit is increased
beyond a safe value by root. E.g. when starting a process after running
"ulimit -H -s unlimited" it will then attempt to use a stack size of the
maximum 1GB which is far too big for metag's limited user virtual
address space (stack_top is usually 0x3ffff000):
BUG: failure at fs/exec.c:589/shift_arg_pages()!
Signed-off-by: James Hogan <james.hogan@imgtec.com> Cc: Helge Deller <deller@gmx.de> Cc: "James E.J. Bottomley" <jejb@parisc-linux.org> Cc: linux-parisc@vger.kernel.org Cc: linux-metag@vger.kernel.org Cc: John David Anglin <dave.anglin@bell.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Volatile access doesn't really imply the compiler barrier. Volatile access
is only ordered with respect to other volatile accesses, it isn't ordered
with respect to general memory accesses. Gcc may reorder memory accesses
around volatile access, as we can see in this simple example (if we
compile it with optimization, both increments of *b will be collapsed to
just one):
void fn(volatile int *a, long *b)
{
(*b)++;
*a = 10;
(*b)++;
}
Consequently, we need the compiler barrier after a write to the volatile
variable, to make sure that the compiler doesn't reorder the volatile
write with something else.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Acked-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: James Hogan <james.hogan@imgtec.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The register CLASS_D_CONTROL_1 is marked as volatile because it contains
a bit, DAC_MUTE, which is also mirrored in the ADC_DAC_CONTROL_1
register. This causes problems for the "Speaker Switch" control, which
will report an error if the CODEC is suspended because it relies on a
volatile register.
To resolve this issue mark CLASS_D_CONTROL_1 as non-volatile and
manually keep the register cache in sync by updating both bits when
changing the mute status.
Reported-by: Shawn Guo <shawn.guo@linaro.org> Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com> Tested-by: Shawn Guo <shawn.guo@linaro.org> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
It hardly could be ever bigger than PAGE_SIZE even for large-scale machine,
but for consistency with its couterpart pcpu_mem_zalloc(),
use pcpu_mem_free() instead.
Commit b4916cb17c26 ("percpu: make pcpu_free_chunk() use
pcpu_mem_free() instead of kfree()") addressed this problem, but
missed this one.
Having multiple windows with the same target and attribute is actually
legal, and can be useful for PCIe windows, when PCIe BARs have a size
that isn't a power of two, and we therefore need to create several
MBus windows to cover the PCIe BAR for a given PCIe interface.
Fixes: fddddb52a6c4 ('bus: introduce an Marvell EBU MBus driver') Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Link: https://lkml.kernel.org/r/1397823593-1932-7-git-send-email-thomas.petazzoni@free-electrons.com Tested-by: Neil Greatorex <neil@fatboyfat.co.uk> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
mvebu_pcie_handle_membase_change() and
mvebu_pcie_handle_iobase_change() do not correctly compute the window
size. PCI uses an inclusive start/end address pair, which requires a
+1 when converting to size.
This only worked because a bug in the mbus driver allowed it to
silently accept and round up bogus sizes.
Fix this by adding one to the computed size.
Fixes: 45361a4fe446 ('PCIe driver for Marvell Armada 370/XP systems') Signed-off-by: Willy Tarreau <w@1wt.eu> Reviewed-By: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Link: https://lkml.kernel.org/r/1397823593-1932-5-git-send-email-thomas.petazzoni@free-electrons.com Tested-by: Neil Greatorex <neil@fatboyfat.co.uk> Acked-by: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
According to the Armada 370 and Armada XP datasheets, the part of the
Device Bus register that configure the bus width should contain 0 for
a 8 bits bus width, and 1 for a 16 bits bus width (other values are
unsupported/reserved).
However, the current conversion done in the driver to convert from a
bus width in bits to the value expected by the register leads to
setting the register to 1 for a 8 bits bus, and 2 for a 16 bits bus.
This mistake was compensated by a mistake in the existing Device Tree
files for Armada 370/XP platforms: they were declaring a 8 bits bus
width, while the hardware in fact uses a 16 bits bus width.
This commit fixes that by adjusting the conversion logic.
This patch fixes a bug that was introduced in 3edad321b1bd2e6c8b5f38146c115c8982438f06 ('drivers: memory: Introduce
Marvell EBU Device Bus driver'), which was merged in v3.11.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Link: https://lkml.kernel.org/r/1397489361-5833-2-git-send-email-thomas.petazzoni@free-electrons.com Fixes: 3edad321b1bd ('drivers: memory: Introduce Marvell EBU Device Bus driver') Acked-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com> Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
After a @pwq is scheduled for emergency execution, other workers may
consume the affectd work items before the rescuer gets to them. This
means that a workqueue many have pwqs queued on @wq->maydays list
while not having any work item pending or in-flight. If
destroy_workqueue() executes in such condition, the rescuer may exit
without emptying @wq->maydays.
This currently doesn't cause any actual harm. destroy_workqueue() can
safely destroy all the involved data structures whether @wq->maydays
is populated or not as nobody access the list once the rescuer exits.
However, this is nasty and makes future development difficult. Let's
update rescuer_thread() so that it empties @wq->maydays after seeing
should_stop to guarantee that the list is empty on rescuer exit.
There is a race condition between rescuer_thread() and
pwq_unbound_release_workfn().
Even after a pwq is scheduled for rescue, the associated work items
may be consumed by any worker. If all of them are consumed before the
rescuer gets to them and the pwq's base ref was put due to attribute
change, the pwq may be released while still being linked on
@wq->maydays list making the rescuer dereference already freed pwq
later.
Make send_mayday() pin the target pwq until the rescuer is done with
it.
The nfsv4 state code has always assumed a one-to-one correspondance
between lock stateid's and lockowners even if it appears not to in some
places.
We may actually change that, but for now when FREE_STATEID releases a
lock stateid it also needs to release the parent lockowner.
Symptoms were a subsequent LOCK crashing in find_lockowner_str when it
calls same_lockowner_ino on a lockowner that unexpectedly has an empty
so_stateids list.
Signed-off-by: J. Bruce Fields <bfields@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Aside from making it clearer what is non-trivial in create_client(), it
also fixes a bug whereby we can call free_client() before idr_init()
has been called.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The recent Intel H97/Z97 chipsets need the similar setups like other
Intel chipsets for snooping, etc. Especially without snooping, the
audio playback stutters or gets corrupted. This fix patch just adds
the corresponding PCI ID entry with the proper flags.
Since commit 1df5a06a ("ALSA: hda - hdmi: Fix programmed active channel
count") channel count is no longer being set if monitor_present is 0.
This is because setting the count was moved after the CA value is
determined, which is only after the monitor_present check in
hdmi_setup_audio_infoframe().
Unfortunately, in some cases, such as with a non-spec-compliant codec or
with a problematic video driver, monitor_present is always 0. As a
specific example, this seems to happen with gen1 ATV (SiI1390 codec),
causing left-channel-only stereo playback (multi-channel playback has
apparently never worked with this codec despite it reporting 8 channels,
reason unknown).
Simply setting converter channel count without setting the pin infoframe
and channel mapping as well does not theoretically make much sense as
this will just mean they are out-of-sync and multichannel playback will
have a wrong channel mapping.
However, adding back just setting the converter channel count even in
no-monitor case is the safest change which at least fixes the stereo
playback regression on SiI1390 codec. Do that.
Signed-off-by: Anssi Hannula <anssi.hannula@iki.fi> Reported-by: Stephan Raue <stephan@openelec.tv> Tested-by: Stephan Raue <stephan@openelec.tv> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The T540p has a touchpad with pnp-id LEN0034, all the models with this
pnp-id have the same min/max values, except the T540p where the values are
slightly off. Fix them to be identical.
This is a preparation patch for simplifying the quirk table.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The hw_version 3 Elantech touchpad on the Gigabyte U2442 does not accept
0x0b as initialization value for r10, this stand-alone version of the
driver: http://planet76.com/drivers/elantech/psmouse-elantech-v6.tar.bz2
Uses 0x03 which does work, so this means not setting bit 3 of r10 which
sets: "Enable Real H/W Resolution In Absolute mode"
Which will result in half the x and y resolution we get with that bit set,
so simply not setting it everywhere is not a solution. We've been unable to
find a way to identify touchpads where setting the bit will fail, so this
patch uses a dmi based blacklist for this.
https://bugzilla.kernel.org/show_bug.cgi?id=61151
Reported-by: Philipp Wolfer <ph.wolfer@gmail.com> Tested-by: Philipp Wolfer <ph.wolfer@gmail.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
After issuing ATKBD_CMD_RESET_DIS, keyboard on some LG laptops stops
working. The workaround is to stop issuing ATKBD_CMD_RESET_DIS commands.
In order to keep changes in atkbd driver to the minimum we check DMI
signature and only skip ATKBD_CMD_RESET_DIS if we are running on LG
LW25-B7HV or P1-J273B.
The replacement of the 'count' variable by two variables 'incs' and
'decs' to resolve some race conditions during module unloading was done
in parallel with some cleanup in the trace subsystem, and was integrated
as a merge.
Unfortunately, the formula for this replacement was wrong in the tracing
code, and the refcount in the traces was not usable as a result.
Use 'count = incs - decs' to compute the user count.
autofs needs to be able to see private data dentry flags for its dentrys
that are being created but not yet hashed and for its dentrys that have
been rmdir()ed but not yet freed. It needs to do this so it can block
processes in these states until a status has been returned to indicate
the given operation is complete.
It does this by keeping two lists, active and expring, of dentrys in
this state and uses ->d_release() to keep them stable while it checks
the reference count to determine if they should be used.
But with the recent lockref changes dentrys being freed sometimes don't
transition to a reference count of 0 before being freed so autofs can
occassionally use a dentry that is invalid which can lead to a panic.
Signed-off-by: Ian Kent <raven@themaw.net> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The crypto algorithm modules utilizing the crypto daemon could
be used early when the system start up. Using module_init
does not guarantee that the daemon's work queue is initialized
when the cypto alorithm depending on crypto_wq starts. It is necessary
to initialize the crypto work queue earlier at the subsystem
init time to make sure that it is initialized
when used.
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
[PATCH v3 1/2] device_cgroup: check if exception removal is allowed
When the device cgroup hierarchy was introduced in bd2953ebbb53 - devcg: propagate local changes down the hierarchy
a specific case was overlooked. Consider the hierarchy bellow:
A default policy: ALLOW, exceptions will deny access
\
B default policy: ALLOW, exceptions will deny access
There's no need to verify when an new exception is added to B because
in this case exceptions will deny access to further devices, which is
always fine. Hierarchy in device cgroup only makes sure B won't have
more access than A.
But when an exception is removed (by writing devices.allow), it isn't
checked if the user is in fact removing an inherited exception from A,
thus giving more access to B.
This shouldn't be allowed and this patch fixes it by making sure to never allow
exceptions in this case to be removed if the exception is partially or fully
present on the parent.
v3: missing '*' in function description
v2: improved log message and formatting fixes
Whenever a device file is opened and checked against current device
cgroup rules, it uses the same function (may_access()) as when a new
exception rule is added by writing devices.{allow,deny}. And in both
cases, the algorithm is the same, doesn't matter the behavior.
First problem is having device access to be considered the same as rule
checking. Consider the following structure:
A (default behavior: allow, exceptions disallow access)
\
B (default behavior: allow, exceptions disallow access)
A new exception is added to B by writing devices.deny:
c 12:34 rw
When checking if that exception is allowed in may_access():
if (dev_cgroup->behavior == DEVCG_DEFAULT_ALLOW) {
if (behavior == DEVCG_DEFAULT_ALLOW) {
/* the exception will deny access to certain devices */
return true;
Which is ok, since B is not getting more privileges than A, it doesn't
matter and the rule is accepted
Now, consider it's a device file open check and the process belongs to
cgroup B. The access will be generated as:
behavior: allow
exception: c 12:34 rw
The very same chunk of code will allow it, even if there's an explicit
exception telling to do otherwise.
c39a2a3018f8 devcg: prepare may_access() for hierarchy support
To solve this problem, the device file open function was split from the
new exception check.
Second problem is how exceptions are processed by may_access(). The
first part of the said function tries to match fully with an existing
exception:
list_for_each_entry_rcu(ex, &dev_cgroup->exceptions, list) {
if ((refex->type & DEV_BLOCK) && !(ex->type & DEV_BLOCK))
continue;
if ((refex->type & DEV_CHAR) && !(ex->type & DEV_CHAR))
continue;
if (ex->major != ~0 && ex->major != refex->major)
continue;
if (ex->minor != ~0 && ex->minor != refex->minor)
continue;
if (refex->access & (~ex->access))
continue;
match = true;
break;
}
That means the new exception should be contained into an existing one to
be considered a match:
New exception Existing match? notes
b 12:34 rwm b 12:34 rwm yes
b 12:34 r b *:34 rw yes
b 12:34 rw b 12:34 w no extra "r"
b *:34 rw b 12:34 rw no too broad "*"
b *:34 rw b *:34 rwm yes
Which is fine in some cases. Consider:
A (default behavior: deny, exceptions allow access)
\
B (default behavior: deny, exceptions allow access)
In this case the full match makes sense, the new exception cannot add
more access than the parent allows
But this doesn't always work, consider:
A (default behavior: allow, exceptions disallow access)
\
B (default behavior: deny, exceptions allow access)
In this case, a new exception in B shouldn't match any of the exceptions
in A, after all you can't allow something that was forbidden by A. But
consider this scenario:
New exception Existing in A match? outcome
b 12:34 rw b 12:34 r no exception is accepted
Because the new exception has "w" as extra, it doesn't match, so it'll
be added to B's exception list.
The same problem can happen during a file access check. Consider a
cgroup with allow as default behavior:
Access Exception match?
b 12:34 rw b 12:34 r no
In this case, the access didn't match any of the exceptions in the
cgroup, which is required since exceptions will disallow access.
To solve this problem, two new functions were created to match an
exception either fully or partially. In the example above, a partial
check will be performed and it'll produce a match since at least
"b 12:34 r" from "b 12:34 rw" access matches.
When probing with DT, we add each LED one at a time. If we find a LED
without a PWM device (because it is not available yet) we fail the
initialisation, unregister previous LEDs, and then by way of managed
resources, we free the structure.
The problem with this is we may have a scheduled and active work_struct
in this structure, and this results in a nasty kernel oops.
We need to cancel this work_struct properly upon cleanup - and the
cleanup we require is the same cleanup as we do when the LED platform
device is removed. Rather than writing this same code three times,
move it into a separate function and use it in all three places.
Fixes: c971ff185f64 ("leds: leds-pwm: Defer led_pwm_set() if PWM can sleep") Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Bryan Wu <cooloney@gmail.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
This should fix a deadlock that has been reported to us where fan_update()
would hold the fan lock and try to grab the alarm_program_lock to reschedule
an update. On an other CPU, the alarm_program_lock would have been taken
before calling fan_update(), leading to a deadlock.
Media force wake get hangs the machine when the system is booted without
displays attached. The assumption is that (at least some versions of)
the firmware has skipped some initialization in that case.
Empirical evidence suggests we need to reset the media force wake
request register in addition to the render one to avoid hangs.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75895 Reported-by: Imre Deak <imre.deak@intel.com> Reported-by: Darren Hart <dvhart@linux.intel.com> Tested-by: Darren Hart <dvhart@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Fixes a regression introduced by 060810d7abaabca "drm/nouveau: fix locking
issues in page flipping paths". chan->cli->mutex is unlocked a second time
in the fail_unreserve path, fix this by moving mutex_unlock down.
There appear to be a crop of new hardware where the vbios is not
available from PROM/PRAMIN, but there is a valid _ROM method in ACPI.
The data read from PCIROM almost invariably contains invalid
instructions (still has the x86 opcodes), which makes this a low-risk
way to try to obtain a valid vbios image.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76475 Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu> Signed-off-by: Ben Skeggs <bskeggs@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Fixes: a53268be0cb9 ('rtlwifi: rtl8192cu: Fix too long disable of IRQs') Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The current .dts for ste-ccu8540 lacks a 'device_type = "memory"' for
its memory node, relying on an old ppc quirk in order to discover its
memory. Fix the data so that all parsing code can handle it correctly.
Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org> Acked-by: Lee Jones <lee.jones@linaro.org> Acked-by: Linus Walleij <linus.walleij@linaro.org> Cc: linux-arm-kernel@lists.infradead.org Cc: devicetree@vger.kernel.org Cc: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Grant Likely <grant.likely@linaro.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The mvebu-devbus driver had a serious bug, which lead to a 8 bits bus
width declared in the Device Tree being considered as a 16 bits bus
width when configuring the hardware.
This bug in mvebu-devbus driver was compensated by a symetric mistake
in the Armada XP OpenBlocks AX3 Device Tree: a 8 bits bus width was
declared, even though the hardware actually has a 16 bits bus width
connection with the NOR flash.
Now that we have fixed the mvebu-devbus driver to behave according to
its Device Tree binding, this commit fixes the problematic Device Tree
files as well.
This bug was introduced in commit a7d4f81821f7eec3175f8e23dd6949c71ab2da43 ('ARM: mvebu: Add support for
NOR flash device on Openblocks AX3 board') which was merged in v3.10.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Link: https://lkml.kernel.org/r/1397489361-5833-5-git-send-email-thomas.petazzoni@free-electrons.com Fixes: a7d4f81821f7 ('ARM: mvebu: Add support for NOR flash device on Openblocks AX3 board') Acked-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com> Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The mvebu-devbus driver had a serious bug, which lead to a 8 bits bus
width declared in the Device Tree being considered as a 16 bits bus
width when configuring the hardware.
This bug in mvebu-devbus driver was compensated by a symetric mistake
in the Armada XP DB Device Tree: a 8 bits bus width was declared, even
though the hardware actually has a 16 bits bus width connection with
the NOR flash.
Now that we have fixed the mvebu-devbus driver to behave according to
its Device Tree binding, this commit fixes the problematic Device Tree
files as well.
This bug was introduced in commit b484ff42df475c5087d614c4d477273e1906bcb9 ('ARM: mvebu: Add support for
NOR flash device on Armada XP-DB board') which was merged in v3.11.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Link: https://lkml.kernel.org/r/1397489361-5833-4-git-send-email-thomas.petazzoni@free-electrons.com Fixes: b484ff42df47 ('ARM: mvebu: Add support for NOR flash device on Armada XP-DB board') Acked-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com> Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The mvebu-devbus driver had a serious bug, which lead to a 8 bits bus
width declared in the Device Tree being considered as a 16 bits bus
width when configuring the hardware.
This bug in mvebu-devbus driver was compensated by a symetric mistake
in the Armada XP GP Device Tree: a 8 bits bus width was declared, even
though the hardware actually has a 16 bits bus width connection with
the NOR flash.
Now that we have fixed the mvebu-devbus driver to behave according to
its Device Tree binding, this commit fixes the problematic Device Tree
files as well.
This bug was introduced in commit da8d1b38356853c37116f9afa29f15648d7fb159 ('ARM: mvebu: Add support for
NOR flash device on Armada XP-GP board') which was merged in v3.10.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Link: https://lkml.kernel.org/r/1397489361-5833-3-git-send-email-thomas.petazzoni@free-electrons.com Fixes: da8d1b383568 ('ARM: mvebu: Add support for NOR flash device on Armada XP-GP board') Acked-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com> Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Commit 54397d85349f
("ARM: kirkwood: Relocate PCIe device tree nodes")
moved the pcie-controller nodes for the Kirkwood SoCs to the mbus
bus node. For some reason, two boards were not properly converted
and have their pci-controller nodes still in the ocp bus node.
As the corresponding SoC pcie-controller does not exist anymore,
it is likely that pcie is broken on those boards since above commit.
Fix it by moving the pcie related nodes to the correct location.
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com> Fixes: 54397d85349f ("ARM: kirkwood: Relocate PCIe device tree nodes") Acked-by: Andrew Lunn <andrew@lunn.ch> Link: https://lkml.kernel.org/r/1398862602-29595-2-git-send-email-sebastian.hesselbarth@gmail.com Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
In commit 4ca2c04085a1caa903e92a5fc0da25362150aac2 ('ARM: orion5x:
Move to ID based window creation'), the mach-orion5x code was changed
to use the new mvebu-mbus API. However, in the process, a mistake was
made on the crypto SRAM window target ID: it should have been 0x9
(verified in the datasheet) and not 0x0.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com> Link: https://lkml.kernel.org/r/1397400006-4315-2-git-send-email-thomas.petazzoni@free-electrons.com Fixes: 4ca2c04085a1 ('ARM: orion5x: Move to ID based window creation') Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
It is possible for "limit - setpoint + 1" to equal zero, after getting
truncated to a 32 bit variable, and resulting in a divide by zero error.
Using the fully 64 bit divide functions avoids this problem. It also
will cause pos_ratio_polynom() to return the correct value when
(setpoint - limit) exceeds 2^32.
Also uninline pos_ratio_polynom, at Andrew's request.
Various filesystems don't bother checking for a NULL ACL in
posix_acl_equiv_mode, and thus can dereference a NULL pointer when it
gets passed one. This usually happens from the NFS server, as the ACL tools
never pass a NULL ACL, but instead of one representing the mode bits.
Instead of adding boilerplat to all filesystems put this check into one place,
which will allow us to remove the check from other filesystems as well later
on.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reported-by: Ben Greear <greearb@candelatech.com> Reported-by: Marco Munderloh <munderl@tnt.uni-hannover.de>, Cc: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Commit 1c2e004183178 introduced an event handler for the encryption key
refresh complete event with the intent of fixing some LE/SMP cases.
However, this event is shared with BR/EDR and there we actually want to
act only on the auth_complete event (which comes after the key refresh).
If we do not do this we may trigger an L2CAP Connect Request too early
and cause the remote side to return a security block error.
The TEAC UD-H01 firmware sends wrong feedback frequency values, thus
causing the PC to send the samples at a wrong rate, which results in
clicks and crackles in the output.
Add a workaround to detect and fix the corruption.
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
[mick37@gmx.de: use sender->udh01_fb_quirk rather than
ep->udh01_fb_quirk in snd_usb_handle_sync_urb()] Reported-and-tested-by: Mick <mick37@gmx.de> Reported-and-tested-by: Andrea Messa <andr.messa@tiscali.it> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Even if the USB-to-ATAPI converter supported multiple LUNs, this
driver would always detect the same physical device or media because
it doesn't use srb->device->lun in any way.
Tested with an Hewlett-Packard CD-Writer Plus 8200e.
Signed-off-by: Daniele Forsi <dforsi@gmail.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Some OHCI controllers from ATI/AMD seem to have difficulty with
"global" USB suspend, that is, suspending an entire USB bus without
setting the suspend feature for each port connected to a device. When
we try to resume the child devices, the controller gives timeout
errors on the unsuspended ports, requiring resets, and can even cause
ohci-hcd to hang; see
This patch fixes the problem by adding a new quirk flag to ohci-hcd.
The flag causes the ohci_rh_suspend() routine to suspend each
unsuspended, enabled port before suspending the root hub. This
effectively converts the "global" suspend to an ordinary root-hub
suspend. There is no need to unsuspend these ports when the root hub
is resumed, because the child devices will be resumed anyway in the
course of a normal system resume ("global" suspend is never used for
runtime PM).
This patch should be applied to all stable kernels which include
commit 0aa2832dd0d9 (USB: use "global suspend" for system sleep on
USB-2 buses) or a backported version thereof.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Reported-by: Peter Münster <pmlists@free.fr> Tested-by: Peter Münster <pmlists@free.fr> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
When using dt resources retrieval (interrupts and reg properties) there is
no predefined order for these resources in the platform dev resource
table. Also don't expect the number of resource to be always 2.
Signed-off-by: Jean-Jacques Hiblot <jjhiblot@traphandler.com> Acked-by: Boris BREZILLON <b.brezillon@overkiz.com> Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Felipe Balbi <balbi@ti.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Per reference manuals of Freescale P1020 and P2020 SoCs, USB controller
present in these SoCs has bit 17 of USBx_CONTROL register marked as
Reserved - there is no PHY_CLK_VALID bit there.
Testing for this bit in ehci_fsl_setup_phy() behaves differently on two
P1020RDB boards available here - on one board test passes and fsl-usb
init succeeds, but on other board test fails, causing fsl-usb init to
fail.
This patch changes ehci_fsl_setup_phy() not to test PHY_CLK_VALID on
controller version 1.6 that (per manual) does not have this bit.
The driver segfaults when the kernel boots with device tree as the
platform data is then not present and the pointer is deferenced without
checking it is not null. This patch introduces such a check avoiding the
crash.
The version of the drm_tegra_submit structure that was merged all the
way back in 3.10 contains a pad field that was originally intended to
properly pad the following __u64 field. Unfortunately it seems like a
different field was dropped during review that caused this padding to
become unnecessary, but the pad field wasn't removed at that time.
One possible side-effect of this is that since the __u64 following the
pad is now no longer properly aligned, the compiler may (or may not)
introduce padding itself, which results in no predictable ABI.
Rectify this by removing the pad field so that all fields are again
naturally aligned. Technically this is breaking existing userspace ABI,
but given that there aren't any (released) userspace drivers that make
use of this yet, the fallout should be minimal.
Testing the update pending bit directly after issuing an
update is nonsense cause depending on the pixel clock the
CRTC needs a bit of time to execute the flip even when we
are in the VBLANK period.
This is just a non invasive patch to solve the problem at
hand, a more complete and cleaner solution should follow
in the next merge window.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=76564
v2: fix source IDs for CRTC2-6
Signed-off-by: Christian König <christian.koenig@amd.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Depending on the SDVO output_flags SDVO may have multiple connectors
linking to the same encoder (in intel_connector->encoder->base).
Only one of those connectors should be active (ie link to the encoder
thru drm_connector->encoder).
If intel_connector_break_all_links() is called from intel_sanitize_crtc()
we may break the crtc connection of an encoder thru an inactive connector
in which case intel_connector_break_all_links() will not be called again
for the active connector if this happens to come later in the list due to:
if (connector->encoder->base.crtc != &crtc->base)
continue;
in intel_sanitize_crtc().
This will however leave the drm_connector->encoder linkage for this
active connector in place. Subsequently this will cause multiple
warnings in intel_connector_check_state() to trigger and the driver
will eventually die in drm_encoder_crtc_ok() (because of crtc == NULL).
To avoid this remove intel_connector_break_all_links() and move its
code to its two calling functions: intel_sanitize_crtc() and
intel_sanitize_encoder().
This allows to implement the link breaking more flexibly matching
the surrounding code: ie. in intel_sanitize_crtc() we can break the
crtc link separatly after the links to the encoders have been
broken which avoids above problem.
drm/i915: read out the modeset hw state at load and resume time
so goes back to the very beginning of the modeset rework.
v2: This patch takes care of the concernes voiced by Chris Wilson
and Daniel Vetter that only breaking links if the drm_connector
is linked to an encoder may miss some links.
v3: move all encoder handling to encoder loop as suggested by
Daniel Vetter.
Signed-off-by: Egbert Eich <eich@suse.de> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Jani Nikula <jani.nikula@linux.intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
If an md array with externally managed metadata (e.g. DDF or IMSM)
is in use, then we should not set safemode==2 at shutdown because:
1/ this is ineffective: user-space need to be involved in any 'safemode' handling,
2/ The safemode management code doesn't cope with safemode==2 on external metadata
and md_check_recover enters an infinite loop.
Even at shutdown, an infinite-looping process can be problematic, so this
could cause shutdown to hang.
switch_hrtimer_base() calls hrtimer_check_target() which ensures that
we do not migrate a timer to a remote cpu if the timer expires before
the current programmed expiry time on that remote cpu.
But __hrtimer_start_range_ns() calls switch_hrtimer_base() before the
new expiry time is set. So the sanity check in hrtimer_check_target()
is operating on stale or even uninitialized data.
Update expiry time before calling switch_hrtimer_base().
If a cpu is idle and starts an hrtimer which is not pinned on that
same cpu, the nohz code might target the timer to a different cpu.
In the case that we switch the cpu base of the timer we already have a
sanity check in place, which determines whether the timer is earlier
than the current leftmost timer on the target cpu. In that case we
enqueue the timer on the current cpu because we cannot reprogram the
clock event device on the target.
If the timers base is already the target CPU we do not have this
sanity check in place so we enqueue the timer as the leftmost timer in
the target cpus rb tree, but we cannot reprogram the clock event
device on the target cpu. So the timer expires late and subsequently
prevents the reprogramming of the target cpu clock event device until
the previously programmed event fires or a timer with an earlier
expiry time gets enqueued on the target cpu itself.
Add the same target check as we have for the switch base case and
start the timer on the current cpu if it would become the leftmost
timer on the target.
If the last hrtimer interrupt detected a hang it sets hang_detected=1
and programs the clock event device with a delay to let the system
make progress.
If hang_detected == 1, we prevent reprogramming of the clock event
device in hrtimer_reprogram() but not in hrtimer_force_reprogram().
This can lead to the following situation:
hrtimer_interrupt()
hang_detected = 1;
program ce device to Xms from now (hang delay)
We have two timers pending:
T1 expires 50ms from now
T2 expires 5s from now
Now T1 gets canceled, which causes hrtimer_force_reprogram() to be
invoked, which in turn programs the clock event device to T2 (5
seconds from now).
Any hrtimer_start after that will not reprogram the hardware due to
hang_detected still being set. So we effectivly block all timers until
the T2 event fires and cleans up the hang situation.
Add a check for hang_detected to hrtimer_force_reprogram() which
prevents the reprogramming of the hang delay in the hardware
timer. The subsequent hrtimer_interrupt will resolve all outstanding
issues.
[ tglx: Rewrote subject and changelog and fixed up the comment in
hrtimer_force_reprogram() ]
Signed-off-by: Stuart Hayes <stuart.w.hayes@gmail.com> Link: http://lkml.kernel.org/r/53602DC6.2060101@gmail.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
When the kernel is built with CONFIG_PREEMPT it is possible to reach a state
when all modules loaded but some driver still stuck in the deferred list
and there is a need for external event to kick the deferred queue to probe
these drivers.
The issue has been observed on embedded systems with CONFIG_PREEMPT enabled,
audio support built as modules and using nfsroot for root filesystem.
The following log fragment shows such sequence when all audio modules
were loaded but the sound card is not present since the machine driver has
failed to probe due to missing dependency during it's probe.
The board is am335x-evmsk (McASP<->tlv320aic3106 codec) with davinci-evm
machine driver:
In the log the machine driver enters it's probe at 12.719969 (this point it
has been removed from the deferred lists). McASP driver already executing
it's probing (since 12.615118).
The machine driver tries to construct the sound card (12.950839) but did
not found one of the components so it fails. After this McASP driver
registers all the ASoC components (the machine driver still in it's probe
function after it failed to construct the card) and the deferred work is
prepared at 13.099026 (note that this time the machine driver is not in the
lists so it is not going to be handled when the work is executing).
Lastly the machine driver exit from it's probe and the core places it to
the deferred list but there will be no other driver going to load and the
deferred queue is not going to be kicked again - till we have external event
like connecting USB stick, etc.
The proposed solution is to try the deferred queue once more when the last
driver is asking for deferring and we had drivers loaded while this last
driver was probing.
This way we can avoid drivers stuck in the deferred queue.
Signed-off-by: Grant Likely <grant.likely@linaro.org> Reviewed-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The datasheet for EMC1413/EMC1414, which is fully compatible to
EMC1403/1404 and uses the same chip identification, references revision
numbers 0x01, 0x03, and 0x04. Accept the full range of revision numbers
from 0x01 to 0x04 to make sure none are missed.
Signed-off-by: Josef Gajdusek <atx@atx.name>
[Guenter Roeck: Updated headline and description] Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Attempts to set the hysteresis value to a temperature below the target
limit fails with "write error: Numerical result out of range" due to
an inverted comparison.
Signed-off-by: Josef Gajdusek <atx@atx.name> Reviewed-by: Jean Delvare <jdelvare@suse.de>
[Guenter Roeck: Updated headline and description] Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
iovec should be reclaimed whenever caller of rw_copy_check_uvector() returns,
but it doesn't hold when failure happens right after aio_setup_vectored_rw().
Fix that in a such way to avoid hairy goto.
Signed-off-by: Leon Yu <chianglungyu@gmail.com> Signed-off-by: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
For handling a free hugepage in memory failure, the race will happen if
another thread hwpoisoned this hugepage concurrently. So we need to
check PageHWPoison instead of !PageHWPoison.
If hwpoison_filter(p) returns true or a race happens, then we need to
unlock_page(hpage).
In arch/arm/kvm/mmu.c:unmap_range(), we end up doing an extra put_page()
on the stage2 pgd which leads to the BUG in put_page_testzero(). This
happens because a pud_huge() test in unmap_range() returns true when it
should always be false with 2-level pages tables used by 64k pages.
This patch removes support for huge puds if 2-level pagetables are
being used.
Signed-off-by: Mark Salter <msalter@redhat.com>
[catalin.marinas@arm.com: removed #ifndef around PUD_SIZE check] Signed-off-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The invalidation is required in order to maintain proper semantics
under CoW conditions. In scenarios where a process clones several
threads, a thread operating on a core whose DTLB entry for a
particular hugepage has not been invalidated, will be reading from
the hugepage that belongs to the forked child process, even after
hugetlb_cow().
The thread will not see the updated page as long as the stale DTLB
entry remains cached, the thread attempts to write into the page,
the child process exits, or the thread gets migrated to a different
processor.
Signed-off-by: Anthony Iliopoulos <anthony.iliopoulos@huawei.com> Link: http://lkml.kernel.org/r/20140514092948.GA17391@server-36.huawei.corp Suggested-by: Shay Goikhman <shay.goikhman@huawei.com> Acked-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
It's critical for split_huge_page() (and migration) to catch and freeze
all PMDs on rmap walk. It gets tricky if there's concurrent fork() or
mremap() since usually we copy/move page table entries on dup_mm() or
move_page_tables() without rmap lock taken. To get it work we rely on
rmap walk order to not miss any entry. We expect to see destination VMA
after source one to work correctly.
But after switching rmap implementation to interval tree it's not always
possible to preserve expected walk order.
It works fine for dup_mm() since new VMA has the same vma_start_pgoff()
/ vma_last_pgoff() and explicitly insert dst VMA after src one with
vma_interval_tree_insert_after().
But on move_vma() destination VMA can be merged into adjacent one and as
result shifted left in interval tree. Fortunately, we can detect the
situation and prevent race with rmap walk by moving page table entries
under rmap lock. See commit 38a76013ad80.
Problem is that we miss the lock when we move transhuge PMD. Most
likely this bug caused the crash[1].
Fixes: 108d6642ad81 ("mm anon rmap: remove anon_vma_moveto_tail") Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Reviewed-by: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Acked-by: Michel Lespinasse <walken@google.com> Cc: Dave Jones <davej@redhat.com> Cc: David Miller <davem@davemloft.net> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
cfg80211 is notified about connection failures by
__cfg80211_connect_result() call. However, this
function currently does not free cfg80211 sme.
This results in hanging connection attempts in some cases
e.g. when mac80211 authentication attempt is denied,
we have this function call:
ieee80211_rx_mgmt_auth() -> cfg80211_rx_mlme_mgmt() ->
cfg80211_process_auth() -> cfg80211_sme_rx_auth() ->
__cfg80211_connect_result()
but cfg80211_sme_free() is never get called.
Fixes: ceca7b712 ("cfg80211: separate internal SME implementation") Signed-off-by: Eliad Peller <eliadx.peller@intel.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
BIT_WORD() truncates rather than rounds, so the loops in
syncpt_thresh_isr() and _host1x_intr_disable_all_syncpt_intrs() use <=
rather than < in an attempt to process the correct number of registers
when rounding of the conversion of count of bits to count of words is
necessary. However, when rounding isn't necessary because the value is
already a multiple of the divisor (as is the case for all values of
nb_pts the code actually sees), this causes one too many registers to
be processed.
Solve this by using and explicit DIV_ROUND_UP() call, rather than
BIT_WORD(), and comparing with < rather than <=.
__dma_tx_complete is not protected against concurrent
call of serial8250_tx_dma. it can lead to circular tail
index corruption or parallel call of serial_tx_dma on the
same data portion.
This patch fixes this issue by holding the port lock.
fixup_user_fault() is used by the futex code when the direct user access
fails, and the futex code wants it to either map in the page in a usable
form or return an error. It relied on handle_mm_fault() to map the
page, and correctly checked the error return from that, but while that
does map the page, it doesn't actually guarantee that the page will be
mapped with sufficient permissions to be then accessed.
So do the appropriate tests of the vma access rights by hand.
[ Side note: arguably handle_mm_fault() could just do that itself, but
we have traditionally done it in the caller, because some callers -
notably get_user_pages() - have been able to access pages even when
they are mapped with PROT_NONE. Maybe we should re-visit that design
decision, but in the meantime this is the minimal patch. ]
Found by Dave Jones running his trinity tool.
Reported-by: Dave Jones <davej@redhat.com> Acked-by: Hugh Dickins <hughd@google.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Add missing clk_put() call to ata_host_activate() failure path.
Sergei says,
"Hm, I have once fixed that (see that *if* (!ret)) but looks like a
later commit 477c87e90853d136b188c50c0e4a93d01cad872e (ARM:
at91/pata: use gpio_is_valid to check the gpio) broke it again. :-(
Would be good if the changelog did mention that..."
Cc: Andrew Victor <linux@maxim.org.za> Cc: Nicolas Ferre <nicolas.ferre@atmel.com> Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com> Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
After hotplugging CPU1 the first call of interrupt handler for CPU1
oneshot timer was called on CPU0 because it fired before setting IRQ
affinity. Affected are SoCs where Multi Core Timer interrupts are
shared (SPI), e.g. Exynos 4210.
During setup of the MCT timers the clock event device should be
registered after setting the affinity for interrupt. This will prevent
starting the timer too early.
Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Kyungmin Park <kyungmin.park@samsung.com> Cc: Marek Szyprowski <m.szyprowski@samsung.com> Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> Cc: Tomasz Figa <t.figa@samsung.com>, Cc: Daniel Lezcano <daniel.lezcano@linaro.org>, Cc: Kukjin Kim <kgene.kim@samsung.com> Cc: linux-arm-kernel@lists.infradead.org, Link: http://lkml.kernel.org/r/20140416143316.299247848@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The starting cpu is not yet in the online mask so irq_set_affinity()
fails which results in per cpu timers for this cpu ending up on some
other online cpu, ususally cpu 0.
Use irq_force_affinity() which disables the online mask check and
makes things work.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Cc: Kyungmin Park <kyungmin.park@samsung.com> Cc: Marek Szyprowski <m.szyprowski@samsung.com> Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> Cc: Tomasz Figa <t.figa@samsung.com>, Cc: Daniel Lezcano <daniel.lezcano@linaro.org>, Cc: Kukjin Kim <kgene.kim@samsung.com> Cc: linux-arm-kernel@lists.infradead.org, Link: http://lkml.kernel.org/r/20140416143316.106665251@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The current implementation of irq_set_affinity() refuses rightfully to
route an interrupt to an offline cpu.
But there is a special case, where this is actually desired. Some of
the ARM SoCs have per cpu timers which require setting the affinity
during cpu startup where the cpu is not yet in the online mask.
If we can't do that, then the local timer interrupt for the about to
become online cpu is routed to some random online cpu.
The developers of the affected machines tried to work around that
issue, but that results in a massive mess in that timer code.
We have a yet unused argument in the set_affinity callbacks of the irq
chips, which I added back then for a similar reason. It was never
required so it got not used. But I'm happy that I never removed it.
That allows us to implement a sane handling of the above scenario. So
the affected SoC drivers can add the required force handling to their
interrupt chip, switch the timer code to irq_force_affinity() and
things just work.
This does not affect any existing user of irq_set_affinity().
Tagged for stable to allow a simple fix of the affected SoC clock
event drivers.
Reported-and-tested-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Kyungmin Park <kyungmin.park@samsung.com> Cc: Marek Szyprowski <m.szyprowski@samsung.com> Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> Cc: Tomasz Figa <t.figa@samsung.com>, Cc: Daniel Lezcano <daniel.lezcano@linaro.org>, Cc: Kukjin Kim <kgene.kim@samsung.com> Cc: linux-arm-kernel@lists.infradead.org, Link: http://lkml.kernel.org/r/20140416143315.717251504@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
To support the affinity setting of per cpu timers in the early startup
of a not yet online cpu, implement the force logic, which disables the
cpu online check.
Tagged for stable to allow a simple fix of the affected SoC clock
event drivers.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> Cc: Kyungmin Park <kyungmin.park@samsung.com> Cc: Marek Szyprowski <m.szyprowski@samsung.com> Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> Cc: Tomasz Figa <t.figa@samsung.com>, Cc: Daniel Lezcano <daniel.lezcano@linaro.org>, Cc: Kukjin Kim <kgene.kim@samsung.com> Cc: linux-arm-kernel@lists.infradead.org, Link: http://lkml.kernel.org/r/20140416143315.916984416@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
[ tries to modify code, but it's RO, and fails!
ftrace_bug() is called]
When this race happens, ftrace_bug() will produces a nasty warning and
all of the function tracing features will be disabled until reboot.
The simple solution is to treate module load the same way the core
kernel is treated at boot. To hardcode the ftrace function modification
of converting calls to mcount into nops. This is done in init/main.c
there's no reason it could not be done in load_module(). This gives
a better control of the changes and doesn't tie the state of the
module to its notifiers as much. Ftrace is special, it needs to be
treated as such.
The reason this would work, is that the ftrace_module_init() would be
called while the module is in MODULE_STATE_UNFORMED, which is ignored
by the set_all_module_text_ro() call.
projects / people / ms / linux.git / log
