putreg() assumes that the tracee is not running and pt_regs_access() can
safely play with its stack. However a killed tracee can return from
ptrace_stop() to the low-level asm code and do RESTORE_REST, this means
that debugger can actually read/modify the kernel stack until the tracee
does SAVE_REST again.
set_task_blockstep() can race with SIGKILL too and in some sense this
race is even worse, the very fact the tracee can be woken up breaks the
logic.
As Linus suggested we can clear TASK_WAKEKILL around the arch_ptrace()
call, this ensures that nobody can ever wakeup the tracee while the
debugger looks at it. Not only this fixes the mentioned problems, we
can do some cleanups/simplifications in arch_ptrace() paths.
Probably ptrace_unfreeze_traced() needs more callers, for example it
makes sense to make the tracee killable for oom-killer before
access_process_vm().
While at it, add the comment into may_ptrace_stop() to explain why
ptrace_stop() still can't rely on SIGKILL and signal_pending_state().
Reported-by: Salman Qazi <sqazi@google.com>
Reported-by: Suleiman Souhlal <suleiman@google.com>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
By popular demand, arch/aarch64 is now known as arch/arm64. However,
uname -m (and indeed the GNU triplet) still use aarch64 as the machine
string.
This patch fixes native builds of both the kernel and perf tools by
updating the relevant Makefiles to munge the output of uname -m and
set the ARCH variable appropriately.
Cc: <stable@vger.kernel.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The kernel's internal definition of ELF_NGREG uses struct pt_regs, which
means that we disagree with userspace on the size of coredumps since
glibc correctly uses the user-visible struct user_pt_regs.
This patch fixes our ELF_NGREG definition to use struct user_pt_regs
and introduces our own ELF_CORE_COPY_REGS to convert between the user
and kernel structure definitions.
Cc: <stable@vger.kernel.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
This patch (as1642) adds an ehci->priv field for private use by EHCI
platform drivers. The space was provided some time ago, but it didn't
have a name.
Until now none of the platform drivers has used this private space,
but that's about to change in the next patch of this series.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch (as1641) fixes a minor bug in ehci-hcd left over from when
the Chipidea driver was converted to the "ehci-hcd is a library"
scheme. The test for whether the Chipidea platform driver is active
should be IS_ENABLED(), not defined().
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Without this, platform drivers e.g. ehci-omap.c will see a
different version of struct ehci_hcd than ehci-hcd.c and
break reference to 'debug_dir' and 'priv' members when
CONFIG_USB_DEBUG is enabled.
Signed-off-by: Roger Quadros <rogerq@ti.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch (as1644) fixes a race that occurs during startup in
uhci-hcd. If the IRQ line is shared with other devices, it's possible
for the handler routine to be called before the data structures are
fully initialized.
The problem is fixed by adding a check to the IRQ handler routine. If
the initialization hasn't finished yet, the routine will return
immediately.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Reported-by: Don Zickus <dzickus@redhat.com>
Tested-by: "Huang, Adrian (ISS Linux TW)" <adrian.huang@hp.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cleanup and preparation for the next change.
signal_wake_up(resume => true) is overused. None of ptrace/jctl callers
actually want to wakeup a TASK_WAKEKILL task, but they can't specify the
necessary mask.
Turn signal_wake_up() into signal_wake_up_state(state), reintroduce
signal_wake_up() as a trivial helper, and add ptrace_signal_wake_up()
which adds __TASK_TRACED.
This way ptrace_signal_wake_up() can work "inside" ptrace_request()
even if the tracee doesn't have the TASK_WAKEKILL bit set.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The driver can also be built as a module so add MODULE_LICENSE for it. In
addition add MODULE_DESCRIPTION as well.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Commit 0bdfe0cb80 (i2c: omap: sanitize
exit path) changed the interrupt handler to exit early and complete
the transfer after the draining IRQ is handled. As a result, the ARDY
may not be cleared properly, and it may cause all future I2C transfers
to timeout with "timeout waiting for bus ready". This is reproducible
at least with N900 when twl4030_gpio makes a long write (> FIFO size)
during the probe (http://marc.info/?l=linux-omap&m=135818882610432&w=2).
The fix is to continue until we get ARDY interrupt that completes the
transfer. Tested with 3.8-rc4 + N900: 20 boots in a row without errors;
without the patch the problem triggers after few reboots.
Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Acked-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
The errata handling function acks wrong interrupt in case of "Arbitration
lost". Fix it.
Discovered during code review, the real impact of the bug is unknown.
Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Reviewed-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
To fix incorrect P-state frequencies which can happen on
some AMD systems f594065faf
"ACPI: Add fixups for AMD P-state figures"
introduced a quirk to obtain the correct values by reading
from AMD specific MSRs.
This did cause a regression when running a kernel using that
quirk under Xen which does (currently) not pass through MSR
reads to the HW. Instead the guest gets a 0 in return.
And this seems to cause a failure to initialize the ondemand
governour (hard to say for sure as all P-states appear to run
at the same frequency).
While this should also be fixed in the hypervisor (to allow
a guest to read that MSR), this patch is intended to work
around the issue in the meantime. In discussion it turned out
that indeed real HW/BIOSes may choose to not set the valid bit
and thus mark the P-state as invalid. So this could be considered
a fix for broken BIOSes that also works around the issue on Xen.
Signed-off-by: Stefan Bader <stefan.bader@canonical.com>
Cc: 3.7+ <stable@vger.kernel.org>
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
OPP pointers cannot be expected to be valid beyond the boundary
of rcu_read_lock and rcu_read_unlock. Unfortunately, the current
exynos4 busfreq driver does not honor the usage constraint and stores
the OPP pointer in struct busfreq_data. This could potentially
become invalid later such as: across devfreq opp change decisions,
resulting in unpredictable behavior.
To fix this, we introduce a busfreq specific busfreq_opp_info
structure which is used to handle OPP information. OPP information
is de-referenced to voltage and frequency pairs as needed into
busfreq_opp_info structure and used as needed.
Signed-off-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
OPP pointers are protected by RCU locks, the pointer validity is
permissible only under the section of rcu_read_lock to rcu_read_unlock
Add documentation to the effect.
Signed-off-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
OPP pointer is RCU protected, hence after finding it, de-reference
also should be protected with the same RCU context else the OPP
pointer may become invalid.
Reported-by: Jack Mitchell <jack@embed.me.uk>
Tested-by: Alexander Holler <holler@ahsoftware.de>
Tested-by: Jack Mitchell <jack@embed.me.uk>
Acked-by: Alexander Holler <holler@ahsoftware.de>
Signed-off-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
OPP pointer is RCU protected, hence after finding it, de-reference
also should be protected with the same RCU context else the OPP
pointer may become invalid.
Reported-by: Alexander Holler <holler@ahsoftware.de>
Tested-by: Alexander Holler <holler@ahsoftware.de>
Acked-by: Alexander Holler <holler@ahsoftware.de>
Signed-off-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When a transition to the D3cold power state is requested,
acpi_device_set_power() first carries out a transition to D3hot and
then turns off the device's power resources. However, it fails to
update the device's power.state field appropriately and D3hot is
stored in it as a result.
Fix this, but make sure that the device's power state will be
D3hot if its power resources cannot be turned off in the final
step.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Make acpi_power_state_string() return "D3cold" as the string
representation of ACPI power state D3cold instead of "D3" returned
currently, which is confusing.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
After the only user of acpi_power_on_resources(),
acpi_bus_init_power(), has been changed to avoid calling it
for state equal to ACPI_STATE_D3_COLD, it doesn't have to special
case that state any more.
For this reason, modify the checks in acpi_power_on_resources()
so that it returns -EINVAL for ACPI_STATE_D3_COLD as it should.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The ACPI specitication (ACPI 5, Sections 7.2.8 - 7.2.11) requires
that the _PSn (n = 0..3) method, if present, be executed after the
power resources for the given device power state have been set
appropriately. However, acpi_device_set_power() does that only
if the new power state is going to be higher-power (lower-number)
than the power state the device is in already. Otherwise, the
ordering is reverse to protect against situations in which _PSn
might access device registers unavailable after configuring the
power resources for power state Dn (D3 meaning D3hot).
Such situations are very unlikely to happen, though, and _PSn may
actually be implemented with the assumption that power resources
have been configured for power state Dn in advance, so change the
code to follow the specification literally.
This change was previously porposed in a different form by Lv Zheng.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
To reduce code duplication between acpi_device_set_power() and
acpi_bus_init_power(), introduce a new helper function for executing
ACPI devices' _PSn (n = 0..3) methods, acpi_dev_pm_explicit_set().
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The ACPI specification requires the _PSC method to be present under
a device object if its power state cannot be inferred from the states
of power resources used by it (ACPI 5, Section 7.6.2). However, it
also requires that (for power states D0-D2 and D3hot) if the _PSn
(n = 0, 1, 2, 3) method is present under the device object, it also
must be executed after the power resources have been set
appropriately for the device to go into power state Dn (D3 means
D3hot in this case). Thus it is not clear from the specification
whether or not the _PSn method should be executed if the initial
configuraion of power resources used by the device indicates power
state Dn and the _PSC method is not present.
The current implementation of acpi_bus_init_power() is based on the
assumption that it should not be necessary to execute _PSn in the
above situation, but experience shows that in fact that assumption
need not be satisfied. For this reason, make acpi_bus_init_power()
always execute _PSn if the initial configuration of device power
resources indicates power state Dn.
Reported-and-tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
sizeof when applied to a pointer typed expression gives the size of
the pointer
The semantic patch that makes this output is available
in scripts/coccinelle/misc/noderef.cocci.
More information about semantic patching is available at
http://coccinelle.lip6.fr/
Signed-off-by: Laurent Navet <laurent.navet@gmail.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
in probe() entry of i2c_driver, set the of node of adapter and
call of_i2c_register_devices to register all i2c_client from
dt child-nodes
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
cmd_err is used to handle error code, so it should not be unsigned.
This fixes the following warning when building with W=1 option:
drivers/i2c/busses/i2c-mxs.c: In function 'mxs_i2c_xfer_msg':
drivers/i2c/busses/i2c-mxs.c:331:19: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Acked-by: Marek Vasut <marex@denx.de>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Cc: stable@kernel.org
This contains a single compilation fix for the CODA driver.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
iQIcBAABCAAGBQJQ/kpBAAoJEPFlmONMx+ezojIP/16dVFJtilf5YIMYeKVjyiwE
6wabDdTZK7IMWvESfUZ3Lw9jT8Ean6F0TXSOrQ3aACKUh8w4tx5syvbrH5zOLC9R
DjsnATY34z4fiMoE2azw+21nVd6RwV8yxZCF6Zn70yJL0oEQuL3vztdnhwhuHxW6
1We87uq7rSALmkXy26oXsrLo/J8wb8V+xlgGXnWQDMtaYVkf/5vNSAXttEkhid5d
mByHfB7CZxZ2jMPNyXFEjWwEqAZGyPWOuTDJnFYYV1p8HMej2eqORapacx6dOUNX
u2Tu+xHLSBEZ64ouHE6E5XxcNYeT1CGkZ8xUL/NzGgGdwySKqA8sFFKduBkFMmWe
Kd5b1aICLpXtAmNaALXe9s2ANiNpLY2MegYBsltQaKs1/WFE/1eJ518fiS3eOJWA
asWR8wmxxXLATaOztmB4UQ8jR22sUuAdFYLDpM9AN9I6JKi3flb7/oxLGE6NEUcH
BX7lopuGifSbYZbHr9rurAkFkF0KGVZtq9xo4Z+dy69IYSrwfYK82PUh5UYaRcoy
dx51TFB4Nhh7XBlFAzMzIsU1hQaVFjZPKRyrAV+1FG16Fr8R+rvAuA3tdfB81Ncs
+g9KT3z0ErcjHAyz2hqeIXRR4AnyKmQWk7QfGDM2k3UyY7q9sedyd2TjfeYf0+Lq
5KjwJSeL/8BtpLujGCMC
=7H1R
-----END PGP SIGNATURE-----
Merge tag 'imx-fixes-rc' of git://git.pengutronix.de/git/imx/linux-2.6 into fixes
From Sascha Hauer:
ARM i.MX fixes for -rc.
This contains a single compilation fix for the CODA driver.
* tag 'imx-fixes-rc' of git://git.pengutronix.de/git/imx/linux-2.6:
[media] coda: Fix build due to iram.h rename
This is calling list_del() inside a loop which is a problem when we try
move to the next item on the list. I've converted it to use the _safe
version. And also, as a cleanup, I've converted it to use
list_for_each_entry instead of list_for_each.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
The caller of start_bidx_of_node() should give proper node offsets which
point only direct node blocks. Otherwise, it is a caller's bug.
This patch adds comments to make it clear.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
If some small bios of dirty node pages are supposed to be issued during the
sequential data writes, there-in well-produced consecutive data bios are able
to be split by the small node bios, resulting in performance degradation.
So, let's collect a number of dirty node pages until reaching a threshold.
And, by default, I set the threshold as 2MB, a segment size.
This improves sequential write performance on i5, 512GB SSD (830 w/ SATA2) as
follows.
Before: 231 MB/s -> After: 255 MB/s
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com>
This patch adds f2fs_bmap operation to the data address space.
This enables f2fs to support swapfile.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
This was added for all the file systems before.
See the following commit.
commit id: 0b173bc4da
[PATCH] mm: kill vma flag VM_CAN_NONLINEAR
This patch moves actual ptes filling for non-linear file mappings
into special vma operation: ->remap_pages().
File system must implement this method to get non-linear mappings support,
if it uses filemap_fault() then generic_file_remap_pages() can be used.
Now device drivers can implement this method and obtain nonlinear vma support."
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
Add __init to functions in init_f2fs_fs for code consistency.
Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
Commit 3fed40cc ("Btrfs: cleanup duplicated division functions"), which
was merged into 3.8-rc1, has introduced a regression by removing logic
that was guarding us against bad user input. Bring it back.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Currently you can just destroy a qgroup even though it is in use by other qgroups
or has qgroups assigned to it. This patch prevents destruction of qgroups unless
they are completely unused. Otherwise destroy will return EBUSY.
Reported-by: Eric Hopper <hopper@omnifarious.org>
Signed-off-by: Arne Jansen <sensille@gmx.net>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
If a qgroup that has still assignments is deleted by the user, the corresponding
relations are left in the tree. This leads to an unmountable filesystem.
With this patch, those relations are simple ignored.
Reported-by: Eric Hopper <hopper@omnifarious.org>
Signed-off-by: Arne Jansen <sensille@gmx.net>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Finally we have a build fix for fsl-mxc-udc UDC driver.
We also have a fix for ep0 maxburst setting on DWC3
which could confuse the HW if we tell it we had way
too many streams on that endpoint when it _has_ to be
only one.
cppi_dma support for MUSB got a fix when running as a
module. By dropping the wrong __init annotation, the
function will be available even when we're modules and
we're done with .init.text section.
Last, but not least, we have a fix on FunctionFS which
was causing a bug on our option parsing algorithm.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=4n68
-----END PGP SIGNATURE-----
Merge tag 'fixes-for-v3.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/balbi/usb into usb-linus
Felipe writes:
usb: fixes for v3.8-rc5
Finally we have a build fix for fsl-mxc-udc UDC driver.
We also have a fix for ep0 maxburst setting on DWC3
which could confuse the HW if we tell it we had way
too many streams on that endpoint when it _has_ to be
only one.
cppi_dma support for MUSB got a fix when running as a
module. By dropping the wrong __init annotation, the
function will be available even when we're modules and
we're done with .init.text section.
Last, but not least, we have a fix on FunctionFS which
was causing a bug on our option parsing algorithm.
Add the UART2 muxing data to the board file (this used to be,
erroneously, done in the bootloader).
Cc: stable <stable@vger.kernel.org> [3.7]
Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
The iterator correctly handles of_node_put() calls.
Remove it before continue'ing the loop.
Without this patch you get the following with
CONFIG_OF_DYNAMIC set:
ERROR: Bad of_node_put() on /ocp/timer@44e31000!
[<c001329c>] (unwind_backtrace+0x0/0xe0) from [<c03dd8f0>] (of_node_release+0x2c/0xa0)!
[<c03dd8f0>] (of_node_release+0x2c/0xa0) from [<c03ddea0>] (of_find_matching_node_and_match+0x78/0x90)!
[<c03ddea0>] (of_find_matching_node_and_match+0x78/0x90) from [<c06d349c>] (omap_get_timer_dt+0x78/0x90)!
[<c06d349c>] (omap_get_timer_dt+0x78/0x90) from [<c06d3664>] (omap_dm_timer_init_one.clone.2+0x34/0x2bc)!
[<c06d3664>] (omap_dm_timer_init_one.clone.2+0x34/0x2bc) from [<c06d3a2c>] (omap2_gptimer_clocksource_init.clone.4+0x24/0xa8)!
[<c06d3a2c>] (omap2_gptimer_clocksource_init.clone.4+0x24/0xa8) from [<c06cca58>] (time_init+0x20/0x30)!
[<c06cca58>] (time_init+0x20/0x30) from [<c06c9690>] (start_kernel+0x1a8/0x2fc)!
Signed-off-by: Pantelis Antoniou <panto@antoniou-consulting.com>
Acked-by: Jon Hunter <jon-hunter@ti.com>
[tony@atomide.com: updated description per Jon]
Signed-off-by: Tony Lindgren <tony@atomide.com>
Otherwise we will get:
WARNING: vmlinux.o(.text+0x1d4f0): Section mismatch in reference from the
function omap_init_ocp2scp() to the function .init.text:omap_device_build()
The function omap_init_ocp2scp() references
the function __init omap_device_build().
This is often because omap_init_ocp2scp lacks a __init
annotation or the annotation of omap_device_build is wrong.
Signed-off-by: Tony Lindgren <tony@atomide.com>
If some other kernel subsystem has a module notifier, and adds a kprobe
to a ftrace mcount point (now that kprobes work on ftrace points),
when the ftrace notifier runs it will fail and disable ftrace, as well
as kprobes that are attached to ftrace points.
Here's the error:
WARNING: at kernel/trace/ftrace.c:1618 ftrace_bug+0x239/0x280()
Hardware name: Bochs
Modules linked in: fat(+) stap_56d28a51b3fe546293ca0700b10bcb29__8059(F) nfsv4 auth_rpcgss nfs dns_resolver fscache xt_nat iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack lockd sunrpc ppdev parport_pc parport microcode virtio_net i2c_piix4 drm_kms_helper ttm drm i2c_core [last unloaded: bid_shared]
Pid: 8068, comm: modprobe Tainted: GF 3.7.0-0.rc8.git0.1.fc19.x86_64 #1
Call Trace:
[<ffffffff8105e70f>] warn_slowpath_common+0x7f/0xc0
[<ffffffff81134106>] ? __probe_kernel_read+0x46/0x70
[<ffffffffa0180000>] ? 0xffffffffa017ffff
[<ffffffffa0180000>] ? 0xffffffffa017ffff
[<ffffffff8105e76a>] warn_slowpath_null+0x1a/0x20
[<ffffffff810fd189>] ftrace_bug+0x239/0x280
[<ffffffff810fd626>] ftrace_process_locs+0x376/0x520
[<ffffffff810fefb7>] ftrace_module_notify+0x47/0x50
[<ffffffff8163912d>] notifier_call_chain+0x4d/0x70
[<ffffffff810882f8>] __blocking_notifier_call_chain+0x58/0x80
[<ffffffff81088336>] blocking_notifier_call_chain+0x16/0x20
[<ffffffff810c2a23>] sys_init_module+0x73/0x220
[<ffffffff8163d719>] system_call_fastpath+0x16/0x1b
---[ end trace 9ef46351e53bbf80 ]---
ftrace failed to modify [<ffffffffa0180000>] init_once+0x0/0x20 [fat]
actual: cc:bb:d2:4b:e1
A kprobe was added to the init_once() function in the fat module on load.
But this happened before ftrace could have touched the code. As ftrace
didn't run yet, the kprobe system had no idea it was a ftrace point and
simply added a breakpoint to the code (0xcc in the cc:bb:d2:4b:e1).
Then when ftrace went to modify the location from a call to mcount/fentry
into a nop, it didn't see a call op, but instead it saw the breakpoint op
and not knowing what to do with it, ftrace shut itself down.
The solution is to simply give the ftrace module notifier the max priority.
This should have been done regardless, as the core code ftrace modification
also happens very early on in boot up. This makes the module modification
closer to core modification.
Link: http://lkml.kernel.org/r/20130107140333.593683061@goodmis.org
Cc: stable@vger.kernel.org
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Reported-by: Frank Ch. Eigler <fche@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Basic test logs are available here:
http://www.pwsan.com/omap/testlogs/prcm_fixes_c_v3.8-rc/20130121073904/
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=RtJt
-----END PGP SIGNATURE-----
Merge tag 'omap-fixes-b-for-v3.8-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/pjw/omap-pending into omap-for-v3.8-rc4/fixes
A few OMAP integration fixes for v3.8-rc, for OMAP4 audio and OMAP2 reboot.
Basic test logs are available here:
http://www.pwsan.com/omap/testlogs/prcm_fixes_c_v3.8-rc/20130121073904/
The init_completion() call does reinit not only the variable carrying
the flag that the completion finished, but also initialized the
waitqueue associated with the completion. On the contrary, the
INIT_COMPLETION() call only reinits the flag.
In case there was anything still stuck in the waitqueue, subsequent call
to init_completion() would be able to create possible race condition. This
patch uses the proper function and moves init_completion() into .probe() call
of the driver, to be issued only once.
Note that such scenario is impossible, since two threads can never enter the
mxs_i2c_xfer_msg(), since whole this section is protected by mutex in I2C core.
This by no means allows this issue to exit though.
Signed-off-by: Marek Vasut <marex@denx.de>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
srcip_matches() previously had code like this:
srcip_matches(..., struct sockaddr *rhs) {
/* ... */
struct sockaddr_in6 *vaddr6 = (struct sockaddr_in6 *) &rhs;
return ipv6_addr_equal(..., &vaddr6->sin6_addr);
}
which interpreted the values on the stack after the 'rhs' pointer as an
ipv6 address. The correct thing to do is to use 'rhs', not '&rhs'.
Signed-off-by: Nickolai Zeldovich <nickolai@csail.mit.edu>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
Oleg Nesterov
Will Deacon
Alan Stern
Roger Quadros
Mika Westerberg
Aaro Koskinen
Stefan Bader
Nishanth Menon
Rafael J. Wysocki
Laurent Navet
Barry Song
Fabio Estevam
Nestor Ovroy
Olof Johansson
Dan Carpenter
Jaegeuk Kim
Namjae Jeon
Ilya Dryomov
Chris Mason
Arne Jansen
Greg Kroah-Hartman
Luciano Coelho
Pantelis Antoniou
Tony Lindgren
Rob Clark
Steven Rostedt
Takashi Iwai
Marek Vasut
Jerry Snitselaar
Dmitry Kasatkin
Nickolai Zeldovich