Add the CPU Partnumbers for the new Arm designs.
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-2-suzuki.poulose@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
Sending a patch to the chrome people resulted in a message by
Collabora's mailer daemon:
550 5.1.1 <enric.balletbo@collabora.com>: Recipient address rejected: User unknown in local recipient table (in reply to RCPT TO command)
So remove Eric from all maintainer entries.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Benson Leung <bleung@chromium.org>
Link: https://lore.kernel.org/r/20211021063210.52503-1-u.kleine-koenig@pengutronix.de
When code running on the VC2 stack causes a nested VC exception, the
handler will not handle it as expected but goes again into the error
path.
The result is that the panic() call happening when the VC exception
was raised in an invalid context is called recursively. Fix this by
checking the interrupted stack too and only call panic if it is not
the VC2 stack.
[ bp: Fixup comment. ]
Fixes: 0786138c78 ("x86/sev-es: Add a Runtime #VC Exception Handler")
Reported-by: Xinyang Ge <xing@microsoft.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20211021080833.30875-3-joro@8bytes.org
The value of STACK_TYPE_EXCEPTION_LAST points to the last _valid_
exception stack. Reflect that in the check done in the
vc_switch_off_ist() function.
Fixes: a13644f3a5 ("x86/entry/64: Add entry code for #VC handler")
Reported-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20211021080833.30875-2-joro@8bytes.org
When a nbd device encounters a writeback error, that error will
get propagated to the bd_inode's wb_err field. Then if this nbd
device's backend is disconnected and another is attached, we will
get back the previous writeback error on fsync, which is unexpected.
To fix it, let's use invalidate_disk() helper to invalidate the
disk on disconnect instead of just setting disk's capacity to zero.
Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210922123711.187-5-xieyongji@bytedance.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The lo->lo_device can't be null if the lo->lo_backing_file is set.
So let's remove the unnecessary bdev checks and the entire bdev
variable in __loop_clr_fd() since the lo->lo_backing_file is already
checked before.
Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210922123711.187-4-xieyongji@bytedance.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Use invalidate_disk() helper to simplify the code for gendisk
invalidation.
Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210922123711.187-3-xieyongji@bytedance.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
To hide internal implementation and simplify some driver code,
this adds a helper to invalidate the gendisk. It will clean the
gendisk's associated buffer/page caches and reset its internal
states.
Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210922123711.187-2-xieyongji@bytedance.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add driver for the PCIe controller of the MT7621 SoC.
[bhelgaas: rename from pci-mt7621.c to pcie-mt7621.c; also rename Kconfig
symbol from PCI_MT7621 to PCIE_MT7621]
Link: https://lore.kernel.org/r/20210922050035.18162-3-sergio.paracuellos@gmail.com
Signed-off-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
'tree-ssa-operands.h' included in 'gcc-common.h' is duplicated.
it's also included at line 56.
Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Ye Guojin <ye.guojin@zte.com.cn>
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20211019082910.998257-1-ye.guojin@zte.com.cn
This plugin has no impact on the resulting binary, is disabled
under COMPILE_TEST, and is not enabled on any builds I'm aware of.
Additionally, given the clarified purpose of GCC plugins in the kernel,
remove cyc_complexity.
Cc: Masahiro Yamada <masahiroy@kernel.org>
Cc: Michal Marek <michal.lkml@markovi.net>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-hardening@vger.kernel.org
Cc: linux-kbuild@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Miguel Ojeda <ojeda@kernel.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20211020173554.38122-3-keescook@chromium.org
GCC plugins should only exist when some compiler feature needs to be
proven but does not exist in either GCC nor Clang. For example, if a
desired feature is already in Clang, it should be added to GCC upstream.
Document this explicitly.
Additionally, mark the plugins with matching upstream GCC features as
removable past their respective GCC versions.
Cc: Masahiro Yamada <masahiroy@kernel.org>
Cc: Michal Marek <michal.lkml@markovi.net>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: James Morris <jmorris@namei.org>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: linux-hardening@vger.kernel.org
Cc: linux-kbuild@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Cc: linux-security-module@vger.kernel.org
Cc: llvm@lists.linux.dev
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Miguel Ojeda <ojeda@kernel.org>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20211020173554.38122-2-keescook@chromium.org
The cooling devices have their cooling device set_cur_state
read-writable all the time in the sysfs directory, thus allowing the
userspace to act on it.
The thermal framework is wrongly used by userspace as a power capping
framework by acting on the cooling device opaque state. This one then
competes with the in-kernel governor decision.
We have seen in out-of-tree kernels, a big number of devices which are
abusely declaring themselves as cooling device just to act on their
power.
The role of the thermal framework is to protect the junction
temperature of the silicon. Letting the userspace to play with a
cooling device is invalid and potentially dangerous.
The powercap framework is the right framework to do power capping and
moreover it deals with the aggregation via the dev pm qos.
As the userspace governor is marked deprecated and about to be
removed, there is no point to keep this file writable also in the
future.
Emit a warning and deprecate the interface.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Link: https://lore.kernel.org/r/20211019163506.2831454-2-daniel.lezcano@linaro.org
The userspace governor is sending temperature when polling is active
and trip point crossed events. Nothing else.
AFAICT, this governor is used with custom kernels making the userspace
governor co-existing with another governor on the same thermal zone
because there was no notification mechanism, implying a hack in the
framework to support this configuration.
The new netlink thermal notification is able to provide more
information than the userspace governor and give the opportunity to
the users of this governor to replace it by a dedicated notification
framework.
The userspace governor will be removed as its usage is no longer
needed.
Add a warning message to tell the userspace governor is deprecated.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Link: https://lore.kernel.org/r/20211019163506.2831454-1-daniel.lezcano@linaro.org
Submitting AP probe/null during suspend can cause unexpected
disconnect on resume because of timeout waiting for ack status:
wlan0: Failed to send nullfunc to AP 11:22:33:44:55:66 after 500ms, disconnecting
This is especially the case when we enter suspend when a scan is
ongoing, indeed, scan is cancelled from __ieee80211_suspend, leading
to a corresponding (aborted) scan complete event, which in turn causes
the submission of an immediate monitor null frame (restart_sta_timer).
The corresponding packet or ack will not be processed before resuming,
causing a timeout & disconnect on resume.
Delay the AP probing when suspending/suspended.
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Link: https://lore.kernel.org/r/1634805927-1113-1-git-send-email-loic.poulain@linaro.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Define LC band which is a draft under IEEE 802.11bb.
Current NL80211_BAND_LC is a placeholder band and
will be more defined IEEE 802.11bb progresses.
Signed-off-by: Srinivasan Raju <srini.raju@purelifi.com>
Link: https://lore.kernel.org/r/20211018100143.7565-2-srini.raju@purelifi.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Split __ieee80211_beacon_get() into a separate function for AP mode
ieee80211_beacon_get_ap().
Also, move the code common to all modes (AP, adhoc and mesh) to
a separate function ieee80211_beacon_get_finish().
Signed-off-by: Aloka Dixit <alokad@codeaurora.org>
Link: https://lore.kernel.org/r/20211006040938.9531-2-alokad@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Get channel number from ies is a common logic, so separate it to a new
function, which could also be used by lower driver.
Signed-off-by: Wen Gong <wgong@codeaurora.org>
Link: https://lore.kernel.org/r/20210930081533.4898-1-wgong@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Cleanups of RISC-V SiFive and Microchip DTSes with dtschema. These are
few minor fixes to make DTSes pass the dtschema, without actual
functional effect.
-----BEGIN PGP SIGNATURE-----
iQJEBAABCgAuFiEE3dJiKD0RGyM7briowTdm5oaLg9cFAmFxLQwQHGtyemtAa2Vy
bmVsLm9yZwAKCRDBN2bmhouD1/UGD/4vFvmJdbVPd2tDu81qBwrBx13NrAGj2jlM
6+ZfOzxCZEFd0e+uvhL39f0rm9G7E/TIHtsAc0vL9+InoyByy1UjkCtgDgBAzqPq
LNmL/QqRsWYivIs7kytnl2Q0b4GS4qflJ9YBe7KHr0apoTFYwIF9pAOCnllG5XG4
0N94CK/ucWJgkE8R11Zt0cVLBp3ZDJ+rRc9Afqhw/kTznzdVNxjL7h0a1J0FlfH2
vFDzP34YdUJt5k0eWeu1ZFBpcN5u45/zE/TtiI8hqsehy2+opuiWocKv2duoJZ2F
X48+Jtw3dahq7KbZPVOnxx/nbEDn4/Bsj/B8Mrqv+wk5BQMRl7L0x2LpbFNJoAB2
TDsTr19dvCi09C6eA8aVM/FSLeeuexeeJ9U4luthHZWHjEgA94lGSkCHyBlKZGzG
ro7y25q0lDNBdhxnYZK5PHRJ7sVnSRbMz27cCVQuG6zlSlC8miwNxvTDdscQS5OS
9sylhsSjvSlkOUNZ9t2SgjcBgX55Qap+TvTIjaOL3rFwwbOzEPPw71pI2OYWy8XD
piE3wc8zZwUuMHa1BjsE9o0Jy3X3vGJyrY76tId93S7JVOpZB/xsEhtm7SRXOcHm
kmM1pDOC6OvY+JRhFEohSVBszLhxH2Qm+v7A8z0Ijt1Ellz6kljKYwQLaSp6KCOK
DvRlr/SstQ==
=4+vY
-----END PGP SIGNATURE-----
gpgsig -----BEGIN PGP SIGNATURE-----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=8DbL
-----END PGP SIGNATURE-----
Merge tag 'riscv-sifive-dt-5.16' of git://gitolite.kernel.org/pub/scm/linux/kernel/git/krzk/linux into for-next
RISC-V DTS changes for v5.16
Cleanups of RISC-V SiFive and Microchip DTSes with dtschema. These are
few minor fixes to make DTSes pass the dtschema, without actual
functional effect.
* tag 'riscv-sifive-dt-5.16' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/krzk/linux:
riscv: dts: sifive: add missing compatible for plic
riscv: dts: microchip: add missing compatibles for clint and plic
riscv: dts: sifive: drop duplicated nodes and properties in sifive
riscv: dts: sifive: fix Unleashed board compatible
riscv: dts: sifive: use only generic JEDEC SPI NOR flash compatible
When updating mmu->pkru_mask, the value can only be added but it isn't
reset in advance. This will make mmu->pkru_mask keep the stale data.
Fix this issue.
Fixes: 2d344105f5 ("KVM, pkeys: introduce pkru_mask to cache conditions")
Signed-off-by: Chenyi Qiang <chenyi.qiang@intel.com>
Message-Id: <20211021071022.1140-1-chenyi.qiang@intel.com>
Reviewed-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We never checked for errors on add_disk() as this function
returned void. Now that this is fixed, use the shiny new
error handling.
Acked-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20211015233028.2167651-10-mcgrof@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We never checked for errors on add_disk() as this function
returned void. Now that this is fixed, use the shiny new
error handling.
Acked-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20211015233028.2167651-9-mcgrof@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We never checked for errors on add_disk() as this function
returned void. Now that this is fixed, use the shiny new
error handling.
ubd_disk_register() never returned an error, so just fix
that now and let the caller handle the error condition.
Reviewed-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20211015233028.2167651-8-mcgrof@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We never checked for errors on add_disk() as this function
returned void. Now that this is fixed, use the shiny new
error handling.
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20211015233028.2167651-7-mcgrof@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We never checked for errors on device_add_disk() as this function
returned void. Now that this is fixed, use the shiny new error
handling. The function xlvbd_alloc_gendisk() typically does the
unwinding on error on allocating the disk and creating the tag,
but since all that error handling was stuffed inside
xlvbd_alloc_gendisk() we must repeat the tag free'ing as well.
We set the info->rq to NULL to ensure blkif_free() doesn't crash
on blk_mq_stop_hw_queues() on device_add_disk() error as the queue
will be long gone by then.
Reviewed-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20211015233028.2167651-6-mcgrof@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We never checked for errors on add_disk() as this function
returned void. Now that this is fixed, use the shiny new
error handling.
This driver doesn't do any unwinding with blk_cleanup_disk()
even on errors after add_disk() and so we follow that
tradition.
Acked-by: Coly Li <colyli@suse.de>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20211015233028.2167651-5-mcgrof@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We never checked for errors on add_disk() as this function
returned void. Now that this is fixed, use the shiny new
error handling.
There are two calls to dm_setup_md_queue() which can fail then,
one on dm_early_create() and we can easily see that the error path
there calls dm_destroy in the error path. The other use case is on
the ioctl table_load case. If that fails userspace needs to call
the DM_DEV_REMOVE_CMD to cleanup the state - similar to any other
failure.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20211015233028.2167651-4-mcgrof@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
coccicheck complains about the use of snprintf() in sysfs show
functions:
WARNING use scnprintf or sprintf
Use sysfs_emit instead of scnprintf or sprintf makes more sense.
Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Ye Guojin <ye.guojin@zte.com.cn>
Link: https://lore.kernel.org/r/20211021064931.1047687-1-ye.guojin@zte.com.cn
Signed-off-by: Jens Axboe <axboe@kernel.dk>
netdev->dev_addr will be const soon.
All callers of wdev_address() can take const already.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20211019162816.1384077-2-kuba@kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Commit 406f42fa0d ("net-next: When a bond have a massive amount
of VLANs...") introduced a rbtree for faster Ethernet address look
up. To maintain netdev->dev_addr in this tree we need to make all
the writes to it got through appropriate helpers.
Convert mac80211 from memcpy(... ETH_ADDR) to eth_hw_addr_set():
@@
expression dev, np;
@@
- memcpy(dev->dev_addr, np, ETH_ALEN)
+ eth_hw_addr_set(dev, np)
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20211019162816.1384077-1-kuba@kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Commit 406f42fa0d ("net-next: When a bond have a massive amount
of VLANs...") introduced a rbtree for faster Ethernet address look
up. To maintain netdev->dev_addr in this tree we need to make all
the writes to it got through appropriate helpers.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20211018235021.1279697-12-kuba@kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
In breaking patch buf memory moved from stack to heap and sizeof(buf)
change from size of actual memory to size of the pointer to the heap.
Fix this by holding a separated variable for allocate size.
Fixes: 01f84f0ed3 ("mac80211: reduce stack usage in debugfs")
Signed-off-by: Mordechay Goodstein <mordechay.goodstein@intel.com>
Link: https://lore.kernel.org/r/20211021163035.b9ae48c06e27.I6a6ed197110eae28cf4f6e38ce36828a7c136337@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
With the introduction of LMH in '36c6581214 ("arm64: dts: qcom:
sdm845: Add support for LMh node")' the CPUfreq gained a reference for
the two interrupts from the LMh. This means that if the LMh driver isn't
enabled CPUfreq will not probe and there's no frequency scaling.
Enable LMh to make CPUfreq functional again on SDM845.
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Tested-By: Steev Klimaszewski <steev@kali.org>
Reviewed-and-Tested-by: Bhupesh Sharma <bhupesh.sharma@linaro.org>
Link: https://lore.kernel.org/r/20211017204036.2761122-1-bjorn.andersson@linaro.org
The Qualcomm prima/pronto drivers consists of a remoteproc driver boot
the combined WiFi and Bluetooth firmware on the related co-processor, a
control driver to load calibration and settings firmware, the wcn36xx
WiFi driver and the SMD based Bluetooth driver.
Enable these in the defconfig to provide Bluetooth and WiFi on devices
such as the Qualcomm Dragonboard 410c.
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Reviewed-by: Vinod Koul <vkoul@kernel.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20211008192840.3593362-1-bjorn.andersson@linaro.org
blk_try_enter_queue() already takes rcu_read_lock/unlock, so we can
avoid the second pair in percpu_ref_tryget_live(), use a newly added
percpu_ref_tryget_live_rcu().
As rcu_read_lock/unlock imply barrier()s, it's pretty noticeable,
especially for for !CONFIG_PREEMPT_RCU (default for some distributions),
where __rcu_read_lock/unlock() are not inlined.
3.20% io_uring [kernel.vmlinux] [k] __rcu_read_unlock
3.05% io_uring [kernel.vmlinux] [k] __rcu_read_lock
2.52% io_uring [kernel.vmlinux] [k] __rcu_read_unlock
2.28% io_uring [kernel.vmlinux] [k] __rcu_read_lock
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/6b11c67ea495ed9d44f067622d852de4a510ce65.1634822969.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
