Using a private template is problematic:
1. We can't assign both a zone and a timeout policy
(zone assigns a conntrack template, so we hit problem 1)
2. Using a template needs to take care of ct refcount, else we'll
eventually free the private template due to ->use underflow.
This patch reworks template policy to instead work with existing conntrack.
As long as such conntrack has not yet been placed into the hash table
(unconfirmed) we can still add the timeout extension.
The only caveat is that we now need to update/correct ct->timeout to
reflect the initial/new state, otherwise the conntrack entry retains the
default 'new' timeout.
Side effect of this change is that setting the policy must
now occur from chains that are evaluated *after* the conntrack lookup
has taken place.
No released kernel contains the timeout policy feature yet, so this change
should be ok.
Changes since v2:
- don't handle 'ct is confirmed case'
- after previous patch, no need to special-case tcp/dccp/sctp timeout
anymore
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
tcp, sctp and dccp trackers re-use the userspace ctnetlink states
to index their timeout arrays, which means timeout[0] is never
used. Copy the 'new' state (syn-sent, dccp-request, ..) to 0 as well
so external users can simply read it off timeouts[0] without need to
differentiate dccp/sctp/tcp and udp/icmp/gre/generic.
The alternative is to map all array accesses to 'i - 1', but that
is a much more intrusive change.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
For some SPI controllers, after each word size (specified by bits_per_word)
transimission, the hardware need some delay to make sure the slave has enough
time to receive the whole data.
So introducing one new 'word_delay' field of struct spi_tansfer for slave
devices to set this inter word delay time.
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
An unfortunate consequence of having a strong typing for the input
values to the SMC call is that it also affects the type of the
return values, limiting r0 to 32 bits and r{1,2,3} to whatever
was passed as an input.
Let's turn everything into "unsigned long", which satisfies the
requirements of both architectures, and allows for the full
range of return values.
Reported-by: Julien Grall <julien.grall@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
- Fix mismatch between SVE registers (Z) and FPSIMD register (V)
- Don't prefix the path for [3] with Linux to stay consistent with
[1] and [2].
Signed-off-by: Julien Grall <julien.grall@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
During power domains initialization we acquire power well references for
power wells in the INIT power domain. The rest of power wells - which
BIOS could have left enabled - we can only acquire references as needed
during display HW readout and so must defer sanitization until then
(also implying that we must always do HW readout to cleanup unused power
wells).
Thus during initialization these latter power wells can have a refcount
of 0 while still being enabled. To avoid the false-positive state
mismatch error this causes remove the check from
intel_power_domains_init_hw() and rely on the state check in
intel_power_domains_enable() which follows the HW readout.
v2:
- Add comment to log and code clarifying how unused power wells get
disabled. (Chris)
Fixes: 6dfc4a8f13 ("drm/i915: Verify power domains after enabling them")
Cc: Chris Wilson <chris@chris-wilson.co.uk>
References: https://bugs.freedesktop.org/show_bug.cgi?id=107411
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20180828122231.14336-1-imre.deak@intel.com
When building building AMSDU from non-linear SKB, we hit a
kernel panic when trying to push the padding to the tail.
Instead, put the padding at the head of the next subframe.
This also fixes the A-MSDU subframes to not have the padding
accounted in the length field and not have pad at all for
the last subframe, both required by the spec.
Fixes: 6e0456b545 ("mac80211: add A-MSDU tx support")
Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Reviewed-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
IEEE 802.11-2016 14.10.8.3 HWMP sequence numbering says:
If it is a target mesh STA, it shall update its own HWMP SN to
maximum (current HWMP SN, target HWMP SN in the PREQ element) + 1
immediately before it generates a PREP element in response to a
PREQ element.
Signed-off-by: Yuan-Chi Pang <fu3mo6goo@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Modify the API to include all ACK frames in average ACK
signal strength reporting, not just ACKs for data frames.
Make exposing the data conditional on implementing the
extended feature flag.
This is how it was really implemented in mac80211, update
the code there to use the new defines and clean up some of
the setting code.
Keep nl80211.h source compatibility by keeping the old names.
Signed-off-by: Balaji Pothunoori <bpothuno@codeaurora.org>
[rewrite commit log, change compatibility to be old=new
instead of the other way around, update kernel-doc,
roll in mac80211 changes, make mac80211 depend on valid
bit instead of HW flag]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This fixes:
[BUG] gpio: gpio-adp5588: A possible sleep-in-atomic-context bug
in adp5588_gpio_write()
[BUG] gpio: gpio-adp5588: A possible sleep-in-atomic-context bug
in adp5588_gpio_direction_input()
Reported-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The pin 327 was supposed to be used as a voltage control line for the
SD card regulator, but the SD card port1 does not support UHS-I. It
only supports 3.3V signaling, hence this pin is pointless.
Just a note about the background. At first, hardware engineers tried
to implement the UHS for this port. Then, they needed to shrink the
silicon die size, and gave up the UHS, but forgot to remove the pin
assignment.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
GPIO hole is present in RZ/G1C SoC. Valid GPIO pins on bank3 are in the
range GP3_0 to GP3_16 and GP3_27 to GP3_29. The GPIO pins between GP3_17
to GP3_26 are unused. Add support for handling unused GPIO's.
Signed-off-by: Biju Das <biju.das@bp.renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The SCHMITT tag was being dumped without a separating space.
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The pin_config_[get|set] functions should return ENOTSUPP if
the requested attribute isn't supported.
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
We are using the generic pin configuration interface so
we can set is_generic.
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Add the pinctrl driver for Meson-G12A SoC which share the similar IP as
the previous Meson-AXG SoC, both use same pinmux ops (register layout).
A new driver is needed here due to the differences in the pins.
Starting from Meson-AXG SoC, the pinctrl controller block use 4
continues register bits to specific the pin mux function, while comparing
to old generation SoC which using variable length register bits for
the pin mux definition. The new design greatly simplify the software model.
For the detail example, one 32bit register can be divided into 8 parts,
each has 4 bits whose value start from 0 - 7, each can describe one pin,
the value 0 is always devoted to GPIO function, while 1 - 7 devoted to
the mux pin function.
Please note, the GPIOE is actually located at AO (always on) bank.
Acked-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Xingyu Chen <xingyu.chen@amlogic.com>
Signed-off-by: Yixun Lan <yixun.lan@amlogic.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Add new compatible name for Amlogic's Meson-G12A pin controllers,
add a dt-binding header file which document the detail pin names.
Acked-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Xingyu Chen <xingyu.chen@amlogic.com>
Signed-off-by: Yixun Lan <yixun.lan@amlogic.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
When requesting a gpio as an interrupt, we should make sure to mux the
pin as the GPIO function and configure it to be an input so that various
functions or output signals don't affect the interrupt state of the pin.
So far, we've relied on pinmux configurations in DT to handle this, but
let's explicitly configure this in the code so that DT implementers
don't have to get this part right.
Cc: Doug Anderson <dianders@chromium.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
We rely on devices to use pinmuxing configurations in DT to select the
GPIO function (function 0) if they're going to use the gpio in GPIO
mode. Let's simplify things for driver authors by implementing
gpio_request_enable() for this pinctrl driver to mux out the GPIO
function when the gpio is use from gpiolib.
Cc: Doug Anderson <dianders@chromium.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The interrupt controller hardware in this pin controller has two status
enable bits. The first "normal" status enable bit enables or disables
the summary interrupt line being raised when a gpio interrupt triggers
and the "raw" status enable bit allows or prevents the hardware from
latching an interrupt into the status register for a gpio interrupt.
Currently we just toggle the "normal" status enable bit in the mask and
unmask ops so that the summary irq interrupt going to the CPU's
interrupt controller doesn't trigger for the masked gpio interrupt.
For a level triggered interrupt, the flow would be as follows: the pin
controller sees the interrupt, latches the status into the status
register, raises the summary irq to the CPU, summary irq handler runs
and calls handle_level_irq(), handle_level_irq() masks and acks the gpio
interrupt, the interrupt handler runs, and finally unmask the interrupt.
When the interrupt handler completes, we expect that the interrupt line
level will go back to the deasserted state so the genirq code can unmask
the interrupt without it triggering again.
If we only mask the interrupt by clearing the "normal" status enable bit
then we'll ack the interrupt but it will continue to show up as pending
in the status register because the raw status bit is enabled, the
hardware hasn't deasserted the line, and thus the asserted state latches
into the status register again. When the hardware deasserts the
interrupt the pin controller still thinks there is a pending unserviced
level interrupt because it latched it earlier. This behavior causes
software to see an extra interrupt for level type interrupts each time
the interrupt is handled.
Let's fix this by clearing the raw status enable bit for level type
interrupts so that the hardware stops latching the status of the
interrupt after we ack it. We don't do this for edge type interrupts
because it seems that toggling the raw status enable bit for edge type
interrupts causes spurious edge interrupts.
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The current mac80211 WDS (4-address mode) can be used to cover most of the
Multi-AP requirements for Data frames per the WFA Multi-AP Specification v1.0.
When configuring AP/STA interfaces in 4-address mode, they are able to function
as fronthaul AP/backhaul STA of Multi-AP device complying below
Tx, Rx requirements except one missing STA Rx requirement added by this patch.
Multi-AP specification section 14.1 describes the following requirements:
Transmitter requirements
------------------------
1. Fronthaul AP
i) When DA!=RA of backhaul STA, must use 4-address format
ii) When DA==RA of backhaul STA, shall use either 3-address
or 4-address format with RA updated with STA MAC
(mac80211 support 4-address format via AP/VLAN interface)
2. Backhaul STA
i) When SA!=TA of backhaul STA, must use 4-address format
ii) When SA==TA of backhaul STA, shall use either 3-address
or 4-address format with RA updated with AP MAC
(mac80211 support 4-address format via use_4addr)
Receiver requirements
---------------------
1. Fronthaul AP
i) When SA!=TA of backhaul STA, must support receiving 4-address
format frames
ii) When SA==TA of backhaul STA, must support receiving both
3-address and 4-address format frames
(mac80211 support both 3-addr & 4-addr via AP/VLAN interface)
2. Backhaul STA
i) When DA!=RA of backhaul STA, must support receiving 4-address
format frames
ii) When DA==RA of backhaul STA, must support receiving both
3-address and 4-address format frames
(mac80211 support only receiving 4-address format via
use_4addr)
This patch addresses the above Rx requirement (ii) for backhaul STA to receive
unicast (DA==RA) 3-address frames in addition to 4-address frames.
The current design doesn't accept 3-address frames when configured in 4-address
mode (use_4addr). Hence add a check to allow 3-address frames when DA==RA of
backhaul STA (adhering to Table 9-26 of IEEE Std 802.11™-2016).
This case was tested with a bridged station interface when associated with
a non-mac80211 based vendor AP implementation using 3-address frames for WDS.
STA was able to support the Multi-AP Rx requirement when DA==RA. No issues,
no loops seen when tested with mac80211 based AP as well.
Verified and confirmed all other Tx and Rx requirements of AP and STA for
Multi-AP respectively. They all work using the current mac80211-WDS design.
Signed-off-by: Sathishkumar Muruganandam <murugana@codeaurora.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
We can quite easily switch banks/ports A and B to use
GPIOLIB_IRQCHIP which is code that will be more careful
about handling interrupt descriptors and use a proper
irqdomain for translating the IRQs. This cuts down some
code in favor of using the implementation inside
gpiolib.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This removes the callback into the gpiolib creating a
circular call to convert between GPIO numbers and IRQs
and pushes the whole business into the driver, just
using an array of IRQ bases for the three IRQ capable
ports.
This way we get rid of including <linux/gpio.h> that
no driver should include.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This simplifies and standardizes the AB IRQ handler by using
the for_each_set_bit() library function.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
In the IRQ-related functions, switch to using the hwirq
and port number found from the current struct gpio_chip *
As the lower 3 bits of the IRQ number is identical to the
lower 3 bits of the GPIO number we can cut some corners.
Call directly into the gpiochip to set up the direction
and read the input instead of using the consumer API.
This enabled us to cut the confusing irq_to_gpio() macro
that is a remnant of the old generic GPIO API as well.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
For setting debounce config we want to write an offset in
a per-gpiochip register, and we know which gpiochip we are
on. Instead of a roundtrip over the IRQ number, figure out
what port we are on for this GPIO chip, then index to the
right register and write the value.
This adds the ep93xx_gpio_port() that finds the port index
from a struct gpio_chip * that we can later exploit to
simplify more code.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The chained irq handler should call chained_irq_enter()
and chained_irq_exit() before/after handling the chained
IRQ.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This is closer to what the variable (per bank) actually
means. We have the .gpio_to_irq() hook registered only
when this is true.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Instead of using a global variable, pass around the
struct gpio_chip * pointer and dereference to the state
container struct ep93xx_gpio as needed, like all other
drivers do.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The subject says it all. Cut down on boilerplate.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
In order to clean up the driver I need to cut a few trees,
sorry, variable names, so I can see the forest, sorry driver
properly.
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
As a preparation for multiplatform support, this ensures
that the ep93xx gpio driver can be built without any of
the platform specific header files. We pass the IRQ numbers
as a resource now, and use the virtual mmio base from the
already existing resource, rather than relying on the
hardwired virtual address from the header file.
Some numbers are now hardcoded that came from macros
in the past, but for all I can tell, the driver already
relied on the specific values.
Cc: arm@kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Alexander Sverdlin <alexander.sverdlin@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
In case of error, the function pcim_iomap() returns NULL pointer not
ERR_PTR(). The IS_ERR() test in the return value check should be
replaced with NULL test.
Fixes: 7740d04d90 ("usb: dwc3: pci: Enable ULPI Refclk on platforms where the firmware does not")
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
This patch fixes an issue that maxpacket size of ep0 is incorrect
for SuperSpeed. Otherwise, CDC NCM class with SuperSpeed doesn't
work correctly on this driver because its control read data size
is more than 64 bytes.
Reported-by: Junki Kato <junki.kato.xk@renesas.com>
Fixes: 746bfe63bb ("usb: gadget: renesas_usb3: add support for Renesas USB3.0 peripheral controller")
Cc: <stable@vger.kernel.org> # v4.5+
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Tested-by: Junki Kato <junki.kato.xk@renesas.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
There is no deallocation of fotg210->ep[i] elements, allocated at
fotg210_udc_probe.
The patch adds deallocation of fotg210->ep array elements and simplifies
error path of fotg210_udc_probe().
Found by Linux Driver Verification project (linuxtesting.org).
Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Commit f16443a034 ("USB: gadgetfs, dummy-hcd, net2280: fix locking
for callbacks") was based on a serious misunderstanding. It
introduced regressions into both the dummy-hcd and net2280 drivers.
The problem in dummy-hcd was fixed by commit 7dbd8f4cab ("USB:
dummy-hcd: Fix erroneous synchronization change"), but the problem in
net2280 remains. Namely: the ->disconnect(), ->suspend(), ->resume(),
and ->reset() callbacks must be invoked without the private lock held;
otherwise a deadlock will occur when the callback routine tries to
interact with the UDC driver.
This patch largely is a reversion of the relevant parts of
f16443a034. It also drops the private lock around the calls to
->suspend() and ->resume() (something the earlier patch forgot to do).
This is safe from races with device interrupts because it occurs
within the interrupt handler.
Finally, the patch changes where the ->disconnect() callback is
invoked when net2280_pullup() turns the pullup off. Rather than
making the callback from within stop_activity() at a time when dropping
the private lock could be unsafe, the callback is moved to a point
after the lock has already been dropped.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Fixes: f16443a034 ("USB: gadgetfs, dummy-hcd, net2280: fix locking for callbacks")
Reported-by: D. Ziesche <dziesche@zes.com>
Tested-by: D. Ziesche <dziesche@zes.com>
CC: <stable@vger.kernel.org>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
An incorrect #ifdef caused a pair of harmless warnings when CONFIG_PM_SLEEP
is disabled:
drivers/usb/dwc3/dwc3-of-simple.c:223:12: error: 'dwc3_of_simple_resume' defined but not used [-Werror=unused-function]
static int dwc3_of_simple_resume(struct device *dev)
^~~~~~~~~~~~~~~~~~~~~
drivers/usb/dwc3/dwc3-of-simple.c:213:12: error: 'dwc3_of_simple_suspend' defined but not used [-Werror=unused-function]
static int dwc3_of_simple_suspend(struct device *dev)
Since the #ifdef method is generally hard to get right, use
a simpler __maybe_unused annotation here to let the compiler
drop the unused functions silently. This also improves
compile-time coverage.
Fixes: 76251db865 ("usb: dwc3: of-simple: reset host controller at suspend/resume")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Instead of open coding logic for reading and writing GPIO lines,
use the generic GPIO library. Also switch to using the spinlock
from the generic GPIO to protect the registers.
Cc: linux-snps-arc@lists.infradead.org
Acked-by: Christian Ruppert <christian.ruppert@alitech.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Create a local struct device *dev helper variable to make the code
easier to read.
Most GPIO drivers use "np" (node pointer) rather than "dn" (device node)
to point to the device tree node. Let's follow this convention.
Cc: linux-snps-arc@lists.infradead.org
Acked-by: Christian Ruppert <christian.ruppert@alitech.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
If the "offload" attribute is used to create an IPsec SA
and the .xdo_dev_state_add() fails, the SA creation fails.
However, if the "offload" attribute is used on a device that
doesn't offer it, the attribute is quietly ignored and the SA
is created without an offload.
Along the same line of that second case, it would be good to
have a way for the device to refuse to offload an SA without
failing the whole SA creation. This patch adds that feature
by allowing the driver to return -EOPNOTSUPP as a signal that
the SA may be fine, it just can't be offloaded.
This allows the user a little more flexibility in requesting
offloads and not needing to know every detail at all times about
each specific NIC when trying to create SAs.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
The pointer 'esph' is defined but is never used hence it is redundant
and canbe removed.
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Fixes the following sparse warning:
net/xfrm/xfrm_interface.c:745:12: warning:
symbol 'xfrmi_get_link_net' was not declared. Should it be static?
Fixes: f203b76d78 ("xfrm: Add virtual xfrm interfaces")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
While recently going over bpf_msg_pull_data(), I noticed three
issues which are fixed in here:
1) When we attempt to find the first scatterlist element (sge)
for the start offset, we add len to the offset before we check
for start < offset + len, whereas it should come after when
we iterate to the next sge to accumulate the offsets. For
example, given a start offset of 12 with a sge length of 8
for the first sge in the list would lead us to determine this
sge as the first sge thinking it covers first 16 bytes where
start is located, whereas start sits in subsequent sges so
we would end up pulling in the wrong data.
2) After figuring out the starting sge, we have a short-cut test
in !msg->sg_copy[i] && bytes <= len. This checks whether it's
not needed to make the page at the sge private where we can
just exit by updating msg->data and msg->data_end. However,
the length test is not fully correct. bytes <= len checks
whether the requested bytes (end - start offsets) fit into the
sge's length. The part that is missing is that start must not
be sge length aligned. Meaning, the start offset into the sge
needs to be accounted as well on top of the requested bytes
as otherwise we can access the sge out of bounds. For example
the sge could have length of 8, our requested bytes could have
length of 8, but at a start offset of 4, so we also would need
to pull in 4 bytes of the next sge, when we jump to the out
label we do set msg->data to sg_virt(&sg[i]) + start - offset
and msg->data_end to msg->data + bytes which would be oob.
3) The subsequent bytes < copy test for finding the last sge has
the same issue as in point 2) but also it tests for less than
rather than less or equal to. Meaning if the sge length is of
8 and requested bytes of 8 while having the start aligned with
the sge, we would unnecessarily go and pull in the next sge as
well to make it private.
Fixes: 015632bb30 ("bpf: sk_msg program helper bpf_sk_msg_pull_data")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
There are no more users left of the p9_idpool; delete it.
Link: http://lkml.kernel.org/r/20180711210225.19730-7-willy@infradead.org
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ron Minnich <rminnich@sandia.gov>
Cc: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Dominique Martinet <dominique.martinet@cea.fr>
Replace the custom batch allocation with a slab. Use an IDR to store
pointers to the active requests instead of an array. We don't try to
handle P9_NOTAG specially; the IDR will happily shrink all the way back
once the TVERSION call has completed.
Link: http://lkml.kernel.org/r/20180711210225.19730-6-willy@infradead.org
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ron Minnich <rminnich@sandia.gov>
Cc: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Dominique Martinet <dominique.martinet@cea.fr>