Improve the performance of GRO receive by splitting flows into
multiple hash chains.
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Manage pending per-NAPI GRO packets via list_head.
Return an SKB pointer from the GRO receive handlers. When GRO receive
handlers return non-NULL, it means that this SKB needs to be completed
at this time and removed from the NAPI queue.
Several operations are greatly simplified by this transformation,
especially timing out the oldest SKB in the list when gro_count
exceeds MAX_GRO_SKBS, and napi_gro_flush() which walks the queue
in reverse order.
Signed-off-by: David S. Miller <davem@davemloft.net>
Jonathan writes:
First set of IIO new device support, features and cleanups in the 4.19 cycle
The usual mixed bunch. Particular good to see is the generic
touch screen driver. Will be interesting to see if this works
for other ADCs without major changes.
Core features
* Channel types
- New position relative channel type primarily for touch screen
sensors to feed the generic touchscreen driver.
New device support
* ad5586
- Add support for the AD5311R DAC.
* Generic touch screen driver as an IIO consumer.
- Note this is in input, but due to dependencies is coming through
the IIO tree.
- Specific support for this added to the at91-sama5d2 ADC.
- Various necessary DT bindings added.
Staging Drops
* ADIS16060 gyro
- A device with a very odd interface that was never cleanly supported.
It's now very difficult to get, so unlikely it'll ever be fixed up.
Cleanups and minor features and fixes
* core
- Fix y2038 timestamp issues now the core support is in place.
* 104-quad-8
- Provide some defines for magic numbers to help readability.
- Fix an off by one error in register selection
* ad7606
- Put in a missing function parameter name in a prototype.
* adis16023
- Use generic sign_extend function rather than local version.
* adis16240
- Use generic sign_extend funciton rather than local version.
* at91-sama5d2
- Drop dependency on HAS_DMA now this is handled elsewhere. Will
improve build test coverage.
- Add oversampling ratio control. Note there is a minor ABI change
here to increase the apparent depth to 14 bits so as to allow
for transparent provision of different oversampling ratios that
drop the actual bit depth to 13 or 12 bits.
* hx711
- Add a MAINTAINERS entry for this device.
* inv_mpu6050
- Replace the timestamp fifo 'special' code with generic timestamp
handling.
- Switch to using local store of timestamp divider rather than rate
as that is more helpful for accurate time measurement.
- Fix an unaligned access that didn't seem to be causing any trouble.
- Use the fifo overflow bit to track the overflow status rather than
a software counter.
- New timestamping mechanism to deal with missed sample interrupts.
* stm32-adc
- Drop HAS_DMA build dependency.
* sun4i-gpadc
- Select REGMAP_IRQ a very rarely hit build issue fix.
Jonathan writes:
First set of IIO fixes for the 4.18 cycle.
* bmp280
- Fix wrong relative humidity unit.
* buffer
- Fix a function signature to match the function.
* inv_mpu6050
- Fix a regression in which older ACPI devices won't have working
interrupts due to lack of information on the interrupt type.
* mma8452
- Don't ignore data ready interrupt when handling interrupts as will
look like an unhandled interrupt.
* tsl2x7x/tsl2772
- Avoid a potential division by zero.
Following the removal of ib_create_flow(), adjust the code to get rid of
ib_destroy_flow() too.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
There are no kernel users of this interface so lets drop it.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Now that all users have been converted to use the version of these APIs
that returns a gid_attr pointer we can delete the old entry points.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
While processing a path record entry in CM messages the associated GID
attribute is now also supplied.
Currently for RoCE a netdevice's net namespace pointer and ifindex are
stored in path record entry. Both of these fields of the netdev can change
anytime while processing CM messages. Additionally storing net namespace
without holding reference will lead to use-after-free crash. Therefore it
is removed. Netdevice information for RoCE is instead provided via
referenced gid attribute in ib_cm requests.
Such a design leads to a situation where the kernel can crash when the net
pointer becomes invalid. However today it is always initialized to
init_net, which cannot become invalid. In order to support processing
packets in any arbitrary namespace of the received packet, it is necessary
to avoid such conditions.
This patch removes the dependency on the net pointer and ifindex; instead
it will rely on SGID attribute which contains a pointer to netdev.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Make the sgid_attr available along with path information to the event
consumer, this allows the consumer to keep using the same GID table entry
as the event is related to.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Hold reference to the the sgid_attr which is used in a cm_id until the
cm_id is destroyed.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
The work completion is inspected to determine what dgid table entry was
used to receieve the packet, produces a sgid_attr that matches and sticks
it in the ah_attr.
All callers of this function are now required to release the ah_attr on
success.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Some bugs reproduce quickly only at high CPU-hotplug rates, so the
rcutorture TREE03 scenario now has only 200 milliseconds spacing between
CPU-hotplug operations. At this rate, the torture-test pair of console
messages per operation becomes a bit voluminous. This commit therefore
converts the torture-test set of "verbose" kernel-boot arguments from
bool to int, and prints the extra console messages only when verbose=2.
The default is still verbose=1.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Joel Fernandes is using SRCU to protect from-idle tracepoints, which
requires notrace variants of srcu_read_lock() and srcu_read_unlock()
in order to avoid problems with tracepoints in lockdep. This commit
therefore adds srcu_read_lock_notrace() and srcu_read_unlock_notrace().
[1] http://lkml.kernel.org/r/20180427042656.190746-1-joelaf@google.com
Reported-by: Joel Fernandes <joelaf@google.com>
Intermittently-reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Kristian Evensen says:
In a project I am involved in, we are running ipsec (Strongswan) on
different mt7621-based routers. Each router is configured as an
initiator and has around ~30 tunnels to different responders (running
on misc. devices). Before the flow cache was removed (kernel 4.9), we
got a combined throughput of around 70Mbit/s for all tunnels on one
router. However, we recently switched to kernel 4.14 (4.14.48), and
the total throughput is somewhere around 57Mbit/s (best-case). I.e., a
drop of around 20%. Reverting the flow cache removal restores, as
expected, performance levels to that of kernel 4.9.
When pcpu xdst exists, it has to be validated first before it can be
used.
A negative hit thus increases cost vs. no-cache.
As number of tunnels increases, hit rate decreases so this pcpu caching
isn't a viable strategy.
Furthermore, the xdst cache also needs to run with BH off, so when
removing this the bh disable/enable pairs can be removed too.
Kristian tested a 4.14.y backport of this change and reported
increased performance:
In our tests, the throughput reduction has been reduced from around -20%
to -5%. We also see that the overall throughput is independent of the
number of tunnels, while before the throughput was reduced as the number
of tunnels increased.
Reported-by: Kristian Evensen <kristian.evensen@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
gcc-8 warns for every single definition of a system call entry
point, e.g.:
include/linux/compat.h:56:18: error: 'compat_sys_rt_sigprocmask' alias between functions of incompatible types 'long int(int, compat_sigset_t *, compat_sigset_t *, compat_size_t)' {aka 'long int(int, struct <anonymous> *, struct <anonymous> *, unsigned int)'} and 'long int(long int, long int, long int, long int)' [-Werror=attribute-alias]
asmlinkage long compat_sys##name(__MAP(x,__SC_DECL,__VA_ARGS__))\
^~~~~~~~~~
include/linux/compat.h:45:2: note: in expansion of macro 'COMPAT_SYSCALL_DEFINEx'
COMPAT_SYSCALL_DEFINEx(4, _##name, __VA_ARGS__)
^~~~~~~~~~~~~~~~~~~~~~
kernel/signal.c:2601:1: note: in expansion of macro 'COMPAT_SYSCALL_DEFINE4'
COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
^~~~~~~~~~~~~~~~~~~~~~
include/linux/compat.h:60:18: note: aliased declaration here
asmlinkage long compat_SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__))\
^~~~~~~~~~
The new warning seems reasonable in principle, but it doesn't
help us here, since we rely on the type mismatch to sanitize the
system call arguments. After I reported this as GCC PR82435, a new
-Wno-attribute-alias option was added that could be used to turn the
warning off globally on the command line, but I'd prefer to do it a
little more fine-grained.
Interestingly, turning a warning off and on again inside of
a single macro doesn't always work, in this case I had to add
an extra statement inbetween and decided to copy the __SC_TEST
one from the native syscall to the compat syscall macro. See
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83256 for more details
about this.
[paul.burton@mips.com:
- Rebase atop current master.
- Split GCC & version arguments to __diag_ignore() in order to match
changes to the preceding patch.
- Add the comment argument to match the preceding patch.]
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82435
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Tested-by: Christophe Leroy <christophe.leroy@c-s.fr>
Tested-by: Stafford Horne <shorne@gmail.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
I have occasionally run into a situation where it would make sense to
control a compiler warning from a source file rather than doing so from
a Makefile using the $(cc-disable-warning, ...) or $(cc-option, ...)
helpers.
The approach here is similar to what glibc uses, using __diag() and
related macros to encapsulate a _Pragma("GCC diagnostic ...") statement
that gets turned into the respective "#pragma GCC diagnostic ..." by
the preprocessor when the macro gets expanded.
Like glibc, I also have an argument to pass the affected compiler
version, but decided to actually evaluate that one. For now, this
supports GCC_4_6, GCC_4_7, GCC_4_8, GCC_4_9, GCC_5, GCC_6, GCC_7,
GCC_8 and GCC_9. Adding support for CLANG_5 and other interesting
versions is straightforward here. GNU compilers starting with gcc-4.2
could support it in principle, but "#pragma GCC diagnostic push"
was only added in gcc-4.6, so it seems simpler to not deal with those
at all. The same versions show a large number of warnings already,
so it seems easier to just leave it at that and not do a more
fine-grained control for them.
The use cases I found so far include:
- turning off the gcc-8 -Wattribute-alias warning inside of the
SYSCALL_DEFINEx() macro without having to do it globally.
- Reducing the build time for a simple re-make after a change,
once we move the warnings from ./Makefile and
./scripts/Makefile.extrawarn into linux/compiler.h
- More control over the warnings based on other configurations,
using preprocessor syntax instead of Makefile syntax. This should make
it easier for the average developer to understand and change things.
- Adding an easy way to turn the W=1 option on unconditionally
for a subdirectory or a specific file. This has been requested
by several developers in the past that want to have their subsystems
W=1 clean.
- Integrating clang better into the build systems. Clang supports
more warnings than GCC, and we probably want to classify them
as default, W=1, W=2 etc, but there are cases in which the
warnings should be classified differently due to excessive false
positives from one or the other compiler.
- Adding a way to turn the default warnings into errors (e.g. using
a new "make E=0" tag) while not also turning the W=1 warnings into
errors.
This patch for now just adds the minimal infrastructure in order to
do the first of the list above. As the #pragma GCC diagnostic
takes precedence over command line options, the next step would be
to convert a lot of the individual Makefiles that set nonstandard
options to use __diag() instead.
[paul.burton@mips.com:
- Rebase atop current master.
- Add __diag_GCC, or more generally __diag_<compiler>, abstraction to
avoid code outside of linux/compiler-gcc.h needing to duplicate
knowledge about different GCC versions.
- Add a comment argument to __diag_{ignore,warn,error} which isn't
used in the expansion of the macros but serves to push people to
document the reason for using them - per feedback from Kees Cook.
- Translate severity to GCC-specific pragmas in linux/compiler-gcc.h
rather than using GCC-specific in linux/compiler_types.h.
- Drop all but GCC 8 macros, since we only need to define macros for
versions that we need to introduce pragmas for, and as of this
series that's just GCC 8.
- Capitalize comments in linux/compiler-gcc.h to match the style of
the rest of the file.
- Line up macro definitions with tabs in linux/compiler-gcc.h.]
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Tested-by: Christophe Leroy <christophe.leroy@c-s.fr>
Tested-by: Stafford Horne <shorne@gmail.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
When a USB device attached to a root-hub port sends a wakeup request
to a sleeping system, we do not report the wakeup event to the PM
core. This is because a system resume involves waking up all
suspended USB ports as quickly as possible; without the normal
USB_RESUME_TIMEOUT delay, the host controller driver doesn't set the
USB_PORT_STAT_C_SUSPEND flag and so usb_port_resume() doesn't realize
that a wakeup request was received.
However, some environments (such as Chrome OS) want to have all wakeup
events reported so they can be ascribed to the appropriate device. To
accommodate these environments, this patch adds a new routine to the
hub driver and a corresponding new HCD method to be used when a root
hub resumes. The HCD method returns a bitmap of ports that have
initiated a wakeup signal but not yet completed resuming. The hub
driver can then report to the PM core that the child devices attached
to these ports initiated a wakeup event.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Suggested-by: Anshuman Gupta <anshuman.gupta@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sometimes memory resource may be overlapping with
SystemMemory Operation Region by design, for example if the
memory region is used as a mailbox for communication with a
firmware in the system. One occasion of such mailboxes is
USB Type-C Connector System Software Interface (UCSI).
With regions like that, it is important that the driver is
able to map the memory with the requirements it has. For
example, the driver should be allowed to map the memory as
non-cached memory. However, if the operation region has been
accessed before the driver has mapped the memory, the memory
has been marked as write-back by the time the driver is
loaded. That means the driver will fail to map the memory
if it expects non-cached memory.
To work around the problem, introducing helper that the
drivers can use to temporarily deactivate (unmap)
SystemMemory Operation Regions that overlap with their
IO memory.
Fixes: 8243edf441 ("usb: typec: ucsi: Add ACPI driver")
Cc: stable@vger.kernel.org
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Detected on the Dell XPS 9365.
The laptop has 2 devices that benefit from the hid-generic auto-unbinding.
When those 2 devices are presented to the userspace, udev loads both wacom and
hid-multitouch. When this happens, the code in __hid_bus_reprobe_drivers() is
called concurrently and the second device gets reprobed twice.
An other bug in the power_supply subsystem prevent to remove the wacom driver
if it just finished its initialization, which basically kills the wacom node.
[jkosina@suse.cz: reformat changelog a bit]
Fixes c17a7476e4 ("HID: core: rewrite the hid-generic automatic unbind")
Cc: stable@vger.kernel.org # v4.17
Tested-by: Mario Limonciello <mario.limonciello@dell.com>
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
With the special case handling for Blackfin and Metag was removed by
commit 94e58e0ac3 ("export.h: remove code for prefixing symbols with
underscore"), VMLINUX_SYMBOL_STR() is now equivalent to __stringify().
Replace the remaining usages to prepare for the entire removal of
VMLINUX_SYMBOL_STR().
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Jessica Yu <jeyu@kernel.org>
Pull networking fixes from David Miller:
1) Fix netpoll OOPS in r8169, from Ville Syrjälä.
2) Fix bpf instruction alignment on powerpc et al., from Eric Dumazet.
3) Don't ignore IFLA_MTU attribute when creating new ipvlan links. From
Xin Long.
4) Fix use after free in AF_PACKET, from Eric Dumazet.
5) Mis-matched RTNL unlock in xen-netfront, from Ross Lagerwall.
6) Fix VSOCK loopback on big-endian, from Claudio Imbrenda.
7) Missing RX buffer offset correction when computing DMA addresses in
mvneta driver, from Antoine Tenart.
8) Fix crashes in DCCP's ccid3_hc_rx_send_feedback, from Eric Dumazet.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (34 commits)
sfc: make function efx_rps_hash_bucket static
strparser: Corrected typo in documentation.
qmi_wwan: add support for the Dell Wireless 5821e module
cxgb4: when disabling dcb set txq dcb priority to 0
net_sched: remove a bogus warning in hfsc
net: dccp: switch rx_tstamp_last_feedback to monotonic clock
net: dccp: avoid crash in ccid3_hc_rx_send_feedback()
net: Remove depends on HAS_DMA in case of platform dependency
MAINTAINERS: Add file patterns for dsa device tree bindings
net: mscc: make sparse happy
net: mvneta: fix the Rx desc DMA address in the Rx path
Documentation: e1000: Fix docs build error
Documentation: e100: Fix docs build error
Documentation: e1000: Use correct heading adornment
Documentation: e100: Use correct heading adornment
ipv6: mcast: fix unsolicited report interval after receiving querys
vhost_net: validate sock before trying to put its fd
VSOCK: fix loopback on big-endian systems
net: ethernet: ti: davinci_cpdma: make function cpdma_desc_pool_create static
xen-netfront: Update features after registering netdev
...
The DT node passed here isn't necessarily an OPP node, as this routine
can also be used for cases where the "required-opps" property is present
directly in the device's node. Rename it.
This also removes a stale comment.
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
linux/iio/buffer-dma.h was not updated to when length was changed to
unsigned int.
Fixes: c043ec1ca5 ("iio:buffer: make length types match kfifo types")
Signed-off-by: Phil Reid <preid@electromag.com.au>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
struct itimerspec is not y2038-safe.
Introduce a new struct __kernel_itimerspec based on the kernel internal
y2038-safe struct itimerspec64.
The definition of struct __kernel_itimerspec includes two struct
__kernel_timespec.
Since struct __kernel_timespec has the same representation in native and
compat modes, so does struct __kernel_itimerspec. This helps have a common
entry point for syscalls using struct __kernel_itimerspec.
New y2038-safe syscalls will use this new type. Since most of the new
syscalls are just an update to the native syscalls with the type update,
place the new definition under CONFIG_64BIT_TIME. This helps architectures
that do not support the above config to keep using the old definition of
struct itimerspec.
Also change the get/put_itimerspec64 to use struct__kernel_itimerspec.
This will help 32 bit architectures to use the new syscalls when
architectures select CONFIG_64BIT_TIME.
Signed-off-by: Deepa Dinamani <deepa.kernel@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: arnd@arndb.de
Cc: viro@zeniv.linux.org.uk
Cc: linux-fsdevel@vger.kernel.org
Cc: linux-api@vger.kernel.org
Cc: y2038@lists.linaro.org
Link: https://lkml.kernel.org/r/20180617051144.29756-2-deepa.kernel@gmail.com
Pull rseq fixes from Thomas Gleixer:
"A pile of rseq related fixups:
- Prevent infinite recursion when delivering SIGSEGV
- Remove the abort of rseq critical section on fork() as syscalls
inside rseq critical sections are explicitely forbidden. So no
point in doing the abort on the child.
- Align the rseq structure on 32 bytes in the ARM selftest code.
- Fix file permissions of the test script"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
rseq: Avoid infinite recursion when delivering SIGSEGV
rseq/cleanup: Do not abort rseq c.s. in child on fork()
rseq/selftests/arm: Align 'struct rseq_cs' on 32 bytes
rseq/selftests: Make run_param_test.sh executable
Pull core fixes from Thomas Gleixner:
"Two tiny fixes:
- Add the missing machine_real_restart() to objtools noreturn list so
it stops complaining
- Fix a trivial comment typo"
* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
kernel.h: Fix a typo in comment
objtool: Add machine_real_restart() to the noreturn list
Pull x86 fixes from Thomas Gleixner:
"A set of fixes for x86:
- Make Xen PV guest deal with speculative store bypass correctly
- Address more fallout from the 5-Level pagetable handling. Undo an
__initdata annotation to avoid section mismatch and malfunction
when post init code would touch the freed variable.
- Handle exception fixup in math_error() before calling notify_die().
The reverse call order incorrectly triggers notify_die() listeners
for soemthing which is handled correctly at the site which issues
the floating point instruction.
- Fix an off by one in the LLC topology calculation on AMD
- Handle non standard memory block sizes gracefully un UV platforms
- Plug a memory leak in the microcode loader
- Sanitize the purgatory build magic
- Add the x86 specific device tree bindings directory to the x86
MAINTAINER file patterns"
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/mm: Fix 'no5lvl' handling
Revert "x86/mm: Mark __pgtable_l5_enabled __initdata"
x86/CPU/AMD: Fix LLC ID bit-shift calculation
MAINTAINERS: Add file patterns for x86 device tree bindings
x86/microcode/intel: Fix memleak in save_microcode_patch()
x86/platform/UV: Add kernel parameter to set memory block size
x86/platform/UV: Use new set memory block size function
x86/platform/UV: Add adjustable set memory block size function
x86/build: Remove unnecessary preparation for purgatory
Revert "kexec/purgatory: Add clean-up for purgatory directory"
x86/xen: Add call of speculative_store_bypass_ht_init() to PV paths
x86: Call fixup_exception() before notify_die() in math_error()
Pull locking fixes from Thomas Gleixner:
"A set of fixes and updates for the locking code:
- Prevent lockdep from updating irq state within its own code and
thereby confusing itself.
- Buid fix for older GCCs which mistreat anonymous unions
- Add a missing lockdep annotation in down_read_non_onwer() which
causes up_read_non_owner() to emit a lockdep splat
- Remove the custom alpha dec_and_lock() implementation which is
incorrect in terms of ordering and use the generic one.
The remaining two commits are not strictly fixes. They provide irqsave
variants of atomic_dec_and_lock() and refcount_dec_and_lock(). These
are required to merge the relevant updates and cleanups into different
maintainer trees for 4.19, so routing them into mainline without
actual users is the sanest approach.
They should have been in -rc1, but last weekend I took the liberty to
just avoid computers in order to regain some mental sanity"
* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
locking/qspinlock: Fix build for anonymous union in older GCC compilers
locking/lockdep: Do not record IRQ state within lockdep code
locking/rwsem: Fix up_read_non_owner() warning with DEBUG_RWSEMS
locking/refcounts: Implement refcount_dec_and_lock_irqsave()
atomic: Add irqsave variant of atomic_dec_and_lock()
alpha: Remove custom dec_and_lock() implementation
Pull irq fixes from Thomas Gleixner:
"A set of fixes mostly for the ARM/GIC world:
- Fix the MSI affinity handling in the ls-scfg irq chip driver so it
updates and uses the effective affinity mask correctly
- Prevent binding LPIs to offline CPUs and respect the Cavium erratum
which requires that LPIs which belong to an offline NUMA node are
not bound to a CPU on a different NUMA node.
- Free only the amount of allocated interrupts in the GIC-V2M driver
instead of trying to free log2(nrirqs).
- Prevent emitting SYNC and VSYNC targetting non existing interrupt
collections in the GIC-V3 ITS driver
- Ensure that the GIV-V3 interrupt redistributor is correctly
reprogrammed on CPU hotplug
- Remove a stale unused helper function"
* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
irqdesc: Delete irq_desc_get_msi_desc()
irqchip/gic-v3-its: Fix reprogramming of redistributors on CPU hotplug
irqchip/gic-v3-its: Only emit VSYNC if targetting a valid collection
irqchip/gic-v3-its: Only emit SYNC if targetting a valid collection
irqchip/gic-v3-its: Don't bind LPI to unavailable NUMA node
irqchip/gic-v2m: Fix SPI release on error path
irqchip/ls-scfg-msi: Fix MSI affinity handling
genirq/debugfs: Add missing IRQCHIP_SUPPORTS_LEVEL_MSI debug
Pull block fixes from Jens Axboe:
- Further timeout fixes. We aren't quite there yet, so expect another
round of fixes for that to completely close some of the IRQ vs
completion races. (Christoph/Bart)
- Set of NVMe fixes from the usual suspects, mostly error handling
- Two off-by-one fixes (Dan)
- Another bdi race fix (Jan)
- Fix nbd reconfigure with NBD_DISCONNECT_ON_CLOSE (Doron)
* tag 'for-linus-20180623' of git://git.kernel.dk/linux-block:
blk-mq: Fix timeout handling in case the timeout handler returns BLK_EH_DONE
bdi: Fix another oops in wb_workfn()
lightnvm: Remove depends on HAS_DMA in case of platform dependency
nvme-pci: limit max IO size and segments to avoid high order allocations
nvme-pci: move nvme_kill_queues to nvme_remove_dead_ctrl
nvme-fc: release io queues to allow fast fail
nbd: Add the nbd NBD_DISCONNECT_ON_CLOSE config flag.
block: sed-opal: Fix a couple off by one bugs
blk-mq-debugfs: Off by one in blk_mq_rq_state_name()
nvmet: reset keep alive timer in controller enable
nvme-rdma: don't override opts->queue_size
nvme-rdma: Fix command completion race at error recovery
nvme-rdma: fix possible free of a non-allocated async event buffer
nvme-rdma: fix possible double free condition when failing to create a controller
Revert "block: Add warning for bi_next not NULL in bio_endio()"
block: fix timeout changes for legacy request drivers
This patch adds support for virtual xfrm interfaces.
Packets that are routed through such an interface
are guaranteed to be IPsec transformed or dropped.
It is a generic virtual interface that ensures IPsec
transformation, no need to know what happens behind
the interface. This means that we can tunnel IPv4 and
IPv6 through the same interface and support all xfrm
modes (tunnel, transport and beet) on it.
Co-developed-by: Lorenzo Colitti <lorenzo@google.com>
Co-developed-by: Benedict Wong <benedictwong@google.com>
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Benedict Wong <benedictwong@google.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Acked-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Benedict Wong <benedictwong@google.com>
Tested-by: Antony Antony <antony@phenome.org>
Reviewed-by: Eyal Birger <eyal.birger@gmail.com>
Add a new flowi_xfrm structure with informations needed to do
a xfrm lookup. At the moment it keeps the informations about
the new xfrm interface id needed to lookup xfrm interfaces
that are introduced with a followup patch. We need this new
lookup key as other possible keys, like the ifindex is
already part of the xfrm selector and used as a key to
enforce the output device after the transformation in the
policy/state lookup.
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Acked-by: Shannon Nelson <shannon.nelson@oracle.com>
Acked-by: Benedict Wong <benedictwong@google.com>
Tested-by: Benedict Wong <benedictwong@google.com>
Tested-by: Antony Antony <antony@phenome.org>
Reviewed-by: Eyal Birger <eyal.birger@gmail.com>
We already support setting an output mark at the xfrm_state,
unfortunately this does not support the input direction and
masking the marks that will be applied to the skb. This change
adds support applying a masked value in both directions.
The existing XFRMA_OUTPUT_MARK number is reused for this purpose
and as it is now bi-directional, it is renamed to XFRMA_SET_MARK.
An additional XFRMA_SET_MARK_MASK attribute is added for setting the
mask. If the attribute mask not provided, it is set to 0xffffffff,
keeping the XFRMA_OUTPUT_MARK existing 'full mask' semantics.
Co-developed-by: Tobias Brunner <tobias@strongswan.org>
Co-developed-by: Eyal Birger <eyal.birger@gmail.com>
Co-developed-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Tobias Brunner <tobias@strongswan.org>
Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Pull xen fixes from Juergen Gross:
"This contains the following fixes/cleanups:
- the removal of a BUG_ON() which wasn't necessary and which could
trigger now due to a recent change
- a correction of a long standing bug happening very rarely in Xen
dom0 when a hypercall buffer from user land was not accessible by
the hypervisor for very short periods of time due to e.g. page
migration or compaction
- usage of EXPORT_SYMBOL_GPL() instead of EXPORT_SYMBOL() in a
Xen-related driver (no breakage possible as using those symbols
without others already exported via EXPORT-SYMBOL_GPL() wouldn't
make any sense)
- a simplification for Xen PVH or Xen ARM guests
- some additional error handling for callers of xenbus_printf()"
* tag 'for-linus-4.18-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
xen: Remove unnecessary BUG_ON from __unbind_from_irq()
xen: add new hypercall buffer mapping device
xen/scsiback: add error handling for xenbus_printf
scsi: xen-scsifront: add error handling for xenbus_printf
xen/grant-table: Export gnttab_{alloc|free}_pages as GPL
xen: add error handling for xenbus_printf
xen: share start flags between PV and PVH
struct net are effectively allocated from order-1 pages on x86,
with one object per slab, meaning that the 13 low order bits
of their addresses are zero.
Once shifted by L1_CACHE_SHIFT, this leaves 7 zero-bits,
meaning that net_hash_mix() does not help spreading
objects on various hash tables.
For example, TCP listen table has 32 buckets, meaning that
all netns use the same bucket for port 80 or port 443.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Drivers such as vc4 don't initialize mode_config.funcs until later in
initialization, but we know they're atomic since they've got the flag
set. This avoids oopsing on dereferencing funcs in the new atomic
methods sanity checks.
I moved the atomic check function down below the core flag check, to
avoid needing a prototype.
Signed-off-by: Eric Anholt <eric@anholt.net>
Fixes: ba1f665f16 ("drm: Add checks for atomic_[duplicate/destroy]_state with atomic drivers")
Link: https://patchwork.freedesktop.org/patch/msgid/20180621195428.17447-1-eric@anholt.net
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
syzbot is reporting NULL pointer dereference at wb_workfn() [1] due to
wb->bdi->dev being NULL. And Dmitry confirmed that wb->state was
WB_shutting_down after wb->bdi->dev became NULL. This indicates that
unregister_bdi() failed to call wb_shutdown() on one of wb objects.
The problem is in cgwb_bdi_unregister() which does cgwb_kill() and thus
drops bdi's reference to wb structures before going through the list of
wbs again and calling wb_shutdown() on each of them. This way the loop
iterating through all wbs can easily miss a wb if that wb has already
passed through cgwb_remove_from_bdi_list() called from wb_shutdown()
from cgwb_release_workfn() and as a result fully shutdown bdi although
wb_workfn() for this wb structure is still running. In fact there are
also other ways cgwb_bdi_unregister() can race with
cgwb_release_workfn() leading e.g. to use-after-free issues:
CPU1 CPU2
cgwb_bdi_unregister()
cgwb_kill(*slot);
cgwb_release()
queue_work(cgwb_release_wq, &wb->release_work);
cgwb_release_workfn()
wb = list_first_entry(&bdi->wb_list, ...)
spin_unlock_irq(&cgwb_lock);
wb_shutdown(wb);
...
kfree_rcu(wb, rcu);
wb_shutdown(wb); -> oops use-after-free
We solve these issues by synchronizing writeback structure shutdown from
cgwb_bdi_unregister() with cgwb_release_workfn() using a new mutex. That
way we also no longer need synchronization using WB_shutting_down as the
mutex provides it for CONFIG_CGROUP_WRITEBACK case and without
CONFIG_CGROUP_WRITEBACK wb_shutdown() can be called only once from
bdi_unregister().
Reported-by: syzbot <syzbot+4a7438e774b21ddd8eca@syzkaller.appspotmail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When delivering a signal to a task that is using rseq, we call into
__rseq_handle_notify_resume() so that the registers pushed in the
sigframe are updated to reflect the state of the restartable sequence
(for example, ensuring that the signal returns to the abort handler if
necessary).
However, if the rseq management fails due to an unrecoverable fault when
accessing userspace or certain combinations of RSEQ_CS_* flags, then we
will attempt to deliver a SIGSEGV. This has the potential for infinite
recursion if the rseq code continuously fails on signal delivery.
Avoid this problem by using force_sigsegv() instead of force_sig(), which
is explicitly designed to reset the SEGV handler to SIG_DFL in the case
of a recursive fault. In doing so, remove rseq_signal_deliver() from the
internal rseq API and have an optional struct ksignal * parameter to
rseq_handle_notify_resume() instead.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: peterz@infradead.org
Cc: paulmck@linux.vnet.ibm.com
Cc: boqun.feng@gmail.com
Link: https://lkml.kernel.org/r/1529664307-983-1-git-send-email-will.deacon@arm.com
This patch adds the sha384 pre-computed 0-length hash so that device
drivers can use it when an hardware engine does not support computing a
hash from a 0 length input.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds the sha512 pre-computed 0-length hash so that device
drivers can use it when an hardware engine does not support computing a
hash from a 0 length input.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
