Keyrings whose name begin with a '.' are special internal keyrings and so
userspace isn't allowed to create keyrings by this name to prevent
shadowing. However, the patch that added the guard didn't fix
KEYCTL_JOIN_SESSION_KEYRING. Not only can that create dot-named keyrings,
it can also subscribe to them as a session keyring if they grant SEARCH
permission to the user.
This, for example, allows a root process to set .builtin_trusted_keys as
its session keyring, at which point it has full access because now the
possessor permissions are added. This permits root to add extra public
keys, thereby bypassing module verification.
This also affects kexec and IMA.
This can be tested by (as root):
keyctl session .builtin_trusted_keys
keyctl add user a a @s
keyctl list @s
dwc_otg: make nak_holdoff work as intended with empty queues
If URBs reading from non-periodic split endpoints were dequeued and
the last transfer from the endpoint was a NAK handshake, the resulting
qh->nak_frame value was stale which would result in unnecessarily long
polling intervals for the first subsequent transfer with a fresh URB.
Fixup qh->nak_frame in dwc_otg_hcd_urb_dequeue and also guard against
a case where a single URB is submitted to the endpoint, a NAK was
received on the transfer immediately prior to receiving data and the
device subsequently resubmits another URB past the qh->nak_frame interval.
Phil Elwell [Wed, 26 Apr 2017 16:28:47 +0000 (17:28 +0100)]
clk: bcm2835: Limit PCM clock to OSC and PLLD_PER
It is unwise to use sources other than the oscillator and PLLD_PER for
the PCM peripheral (and perhaps others - TBD) because their rate can
change and they may even be switched off, so explicitly restrict the
choice using dummy entries in the list of potential parents (item index
is significant).
James Hughes [Tue, 25 Apr 2017 09:15:06 +0000 (10:15 +0100)]
brcmfmac: Make skb header writable before use
The driver was making changes to the skb_header without
ensuring it was writable (i.e. uncloned).
This patch also removes some boiler plate header size
checking/adjustment code as that is also handled by the
skb_cow_header function used to make header writable.
Signed-off-by: James Hughes <james.hughes@raspberrypi.org> Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com> Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
James Hughes [Mon, 24 Apr 2017 11:40:50 +0000 (12:40 +0100)]
brcmfmac: Ensure pointer correctly set if skb data location changes
The incoming skb header may be resized if header space is
insufficient, which might change the data adddress in the skb.
Ensure that a cached pointer to that data is correctly set by
moving assignment to after any possible changes.
Signed-off-by: James Hughes <james.hughes@raspberrypi.org> Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com> Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
James Hughes [Wed, 19 Apr 2017 10:13:40 +0000 (11:13 +0100)]
smsc95xx: Use skb_cow_head to deal with cloned skbs
The driver was failing to check that the SKB wasn't cloned
before adding checksum data.
Replace existing handling to extend/copy the header buffer
with skb_cow_head.
Signed-off-by: James Hughes <james.hughes@raspberrypi.org> Acked-by: Eric Dumazet <edumazet@google.com> Acked-by: Woojung Huh <Woojung.Huh@microchip.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Wed, 19 Apr 2017 16:59:21 +0000 (09:59 -0700)]
smsc75xx: use skb_cow_head() to deal with cloned skbs
We need to ensure there is enough headroom to push extra header,
but we also need to check if we are allowed to change headers.
skb_cow_head() is the proper helper to deal with this.
Fixes: d0cad871703b ("smsc75xx: SMSC LAN75xx USB gigabit ethernet adapter driver") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: James Hughes <james.hughes@raspberrypi.org> Signed-off-by: David S. Miller <davem@davemloft.net>
commit 2dcab5984841 ("sctp: avoid BUG_ON on sctp_wait_for_sndbuf")
attempted to avoid a BUG_ON call when the association being used for a
sendmsg() is blocked waiting for more sndbuf and another thread did a
peeloff operation on such asoc, moving it to another socket.
As Ben Hutchings noticed, then in such case it would return without
locking back the socket and would cause two unlocks in a row.
Further analysis also revealed that it could allow a double free if the
application managed to peeloff the asoc that is created during the
sendmsg call, because then sctp_sendmsg() would try to free the asoc
that was created only for that call.
This patch takes another approach. It will deny the peeloff operation
if there is a thread sleeping on the asoc, so this situation doesn't
exist anymore. This avoids the issues described above and also honors
the syscalls that are already being handled (it can be multiple sendmsg
calls).
Joint work with Xin Long.
Fixes: 2dcab5984841 ("sctp: avoid BUG_ON on sctp_wait_for_sndbuf") Cc: Alexander Popov <alex.popov@linux.com> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The protocol field is checked when deleting IPv4 routes, but ignored for
IPv6, which causes problems with routing daemons accidentally deleting
externally set routes (observed by multiple bird6 users).
This can be verified using `ip -6 route del <prefix> proto something`.
Signed-off-by: Mantas Mikulėnas <grawity@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
put_chars() stuffs the buffer it gets into an sg, but that buffer may be
on the stack. This breaks with CONFIG_VMAP_STACK=y (for me, it
manifested as printks getting turned into NUL bytes).
Signed-off-by: Omar Sandoval <osandov@fb.com> Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Amit Shah <amit.shah@redhat.com> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Brad Spengler <spender@grsecurity.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 17ce039b4e54 ("[media] cxusb: don't do DMA on stack")
added a kmalloc'ed bounce buffer for writes, but missed to do the same
for reads. As the read only happens after the write is finished, we can
reuse the same buffer.
As dvb_usb_generic_rw handles a read length of 0 by itself, avoid calling
it using the dvb_usb_generic_read wrapper function.
The buffer allocation for the firmware data was changed in
commit 43fab9793c1f ("[media] dvb-usb: don't use stack for firmware load")
but the same applies for the reset value.
Fixes: 43fab9793c1f ("[media] dvb-usb: don't use stack for firmware load") Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Brad Spengler <spender@grsecurity.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
As reported by Marc Duponcheel <marc@offline.be>, firmware load on
dvb-usb is using the stack, with is not allowed anymore on default
Kernel configurations:
[ 1025.958836] dvb-usb: found a 'WideView WT-220U PenType Receiver (based on ZL353)' in cold state, will try to load a firmware
[ 1025.958853] dvb-usb: downloading firmware from file 'dvb-usb-wt220u-zl0353-01.fw'
[ 1025.958855] dvb-usb: could not stop the USB controller CPU.
[ 1025.958856] dvb-usb: error while transferring firmware (transferred size: -11, block size: 3)
[ 1025.958856] dvb-usb: firmware download failed at 8 with -22
[ 1025.958867] usbcore: registered new interface driver dvb_usb_dtt200u
Under CONFIG_STRICT_DEVMEM, reading System RAM through /dev/mem is
disallowed. However, on x86, the first 1MB was always allowed for BIOS
and similar things, regardless of it actually being System RAM. It was
possible for heap to end up getting allocated in low 1MB RAM, and then
read by things like x86info or dd, which would trip hardened usercopy:
This changes the x86 exception for the low 1MB by reading back zeros for
System RAM areas instead of blindly allowing them. More work is needed to
extend this to mmap, but currently mmap doesn't go through usercopy, so
hardened usercopy won't Oops the kernel.
Reported-by: Tommi Rantala <tommi.t.rantala@nokia.com> Tested-by: Tommi Rantala <tommi.t.rantala@nokia.com> Signed-off-by: Kees Cook <keescook@chromium.org> Cc: Brad Spengler <spender@grsecurity.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Accessing the registers of the RTC block on Tegra requires the module
clock to be enabled. This only works because the RTC module clock will
be enabled by default during early boot. However, because the clock is
unused, the CCF will disable it at late_init time. This causes the RTC
to become unusable afterwards. This can easily be reproduced by trying
to use the RTC:
$ hwclock --rtc /dev/rtc1
This will hang the system. I ran into this by following up on a report
by Martin Michlmayr that reboot wasn't working on Tegra210 systems. It
turns out that the rtc-tegra driver's ->shutdown() implementation will
hang the CPU, because of the disabled clock, before the system can be
rebooted.
What confused me for a while is that the same driver is used on prior
Tegra generations where the hang can not be observed. However, as Peter
De Schrijver pointed out, this is because on 32-bit Tegra chips the RTC
clock is enabled by the tegra20_timer.c clocksource driver, which uses
the RTC to provide a persistent clock. This code is never enabled on
64-bit Tegra because the persistent clock infrastructure does not exist
on 64-bit ARM.
The proper fix for this is to add proper clock handling to the RTC
driver in order to ensure that the clock is enabled when the driver
requires it. All device trees contain the clock already, therefore
no additional changes are required.
Reported-by: Martin Michlmayr <tbm@cyrius.com>
Acked-By Peter De Schrijver <pdeschrijver@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com> Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
[bwh: Backported to 4.9: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When GPE is not enabled, it is not efficient to use the wait polling mode
as it introduces an unexpected scheduler delay.
So before the GPE handler is installed, this patch uses busy polling mode
for all EC(s) and the logic can be applied to non boot EC(s) during the
suspend/resume process.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=191561 Tested-by: Jakobus Schurz <jakobus.schurz@gmail.com> Tested-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The code generating the warning in validate_apic_and_package_id() matches
cpu_data(cpu).apicid (initialized in init_intel()->
detect_extended_topology() using cpuid) against the apicid returned from
xen_apic_read(). Now, xen_apic_read() makes a hypercall to retrieve apicid
for the boot cpu but returns 0 otherwise. Hence the warning gets thrown
for all but the boot cpu.
The idea behind xen_apic_read() returning 0 for apicid is that the
guests (even Dom0) should not need to know what physical processor their
vcpus are running on. This is because we currently do not have topology
information in Xen and also because xen allows more vcpus than physical
processors. However, boot cpu's apicid is required for loading
xen-acpi-processor driver on AMD machines. Look at following patch for
details:
commit 558daa289a40 ("xen/apic: Return the APIC ID (and version) for CPU
0.")
So to get rid of the warning, this patch modifies
xen_cpu_present_to_apicid() to return cpu_data(cpu).apicid instead of
calling xen_apic_read().
The warning is not seen on AMD machines because init_amd() populates
cpu_data(cpu).apicid by calling hard_smp_processor_id()->xen_apic_read()
as opposed to using apicid from cpuid as is done on Intel machines.
Signed-off-by: Mohit Gambhir <mohit.gambhir@oracle.com> Reviewed-by: Juergen Gross <jgross@suse.com> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The accelerometer event relies on the ACERWMID_EVENT_GUID notify.
So, this patch changes the codes to setup accelerometer input device
when detected ACERWMID_EVENT_GUID. It avoids that the accel input
device created on every Acer machines.
In addition, patch adds a clearly parsing logic of accelerometer hid
to acer_wmi_get_handle_cb callback function. It is positive matching
the "SENR" name with "BST0001" device to avoid non-supported hardware.
Reported-by: Bjørn Mork <bjorn@mork.no> Cc: Darren Hart <dvhart@infradead.org> Signed-off-by: Lee, Chun-Yi <jlee@suse.com>
[andy: slightly massage commit message] Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Select DW_DMAC_CORE like the rest of glue drivers do, e.g.
drivers/dma/dw/Kconfig.
While here group selectors under SND_SOC_INTEL_HASWELL and
SND_SOC_INTEL_BAYTRAIL.
Make platforms, which are using a common SST firmware driver, to be
dependent on DMADEVICES.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Acked-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Signed-off-by: Mark Brown <broonie@kernel.org> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If we have large devices (say like the 40t drive I was trying to test with) we
will end up overflowing the int arguments to nbd_set_size and not get the right
size for our device. Fix this by using loff_t everywhere so I don't have to
think about this again. Thanks,
Signed-off-by: Josef Bacik <jbacik@fb.com> Signed-off-by: Jens Axboe <axboe@fb.com>
[bwh: Backported to 4.9: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Creating a lot of cgroups at the same time might stall all worker
threads with kmem cache creation works, because kmem cache creation is
done with the slab_mutex held. The problem was amplified by commits 801faf0db894 ("mm/slab: lockless decision to grow cache") in case of
SLAB and 81ae6d03952c ("mm/slub.c: replace kick_all_cpus_sync() with
synchronize_sched() in kmem_cache_shrink()") in case of SLUB, which
increased the maximal time the slab_mutex can be held.
To prevent that from happening, let's use a special ordered single
threaded workqueue for kmem cache creation. This shouldn't introduce
any functional changes regarding how kmem caches are created, as the
work function holds the global slab_mutex during its whole runtime
anyway, making it impossible to run more than one work at a time. By
using a single threaded workqueue, we just avoid creating a thread per
each work. Ordering is required to avoid a situation when a cgroup's
work is put off indefinitely because there are other cgroups to serve,
in other words to guarantee fairness.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=172981 Link: http://lkml.kernel.org/r/20161004131417.GC1862@esperanza Signed-off-by: Vladimir Davydov <vdavydov.dev@gmail.com> Reported-by: Doug Smythies <dsmythies@telus.net> Acked-by: Michal Hocko <mhocko@suse.com> Cc: Christoph Lameter <cl@linux.com> Cc: David Rientjes <rientjes@google.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Pekka Enberg <penberg@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We've fixed the race condition problem in calculating ext4 checksum
value in commit b47820edd163 ("ext4: avoid modifying checksum fields
directly during checksum veficationon"). However, by this change,
when calculating the checksum value of inode whose i_extra_size is
less than 4, we couldn't calculate the checksum value in a proper way.
This problem was found and reported by Nix, Thank you.
Reported-by: Nix <nix@esperi.org.uk> Signed-off-by: Daeho Jeong <daeho.jeong@samsung.com> Signed-off-by: Youngjin Gil <youngjin.gil@samsung.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
I ran into a stack frame size warning because of the on-stack copy of
the USB device structure:
drivers/media/usb/dvb-usb-v2/dvb_usb_core.c: In function 'dvb_usbv2_disconnect':
drivers/media/usb/dvb-usb-v2/dvb_usb_core.c:1029:1: error: the frame size of 1104 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
Copying a device structure like this is wrong for a number of other reasons
too aside from the possible stack overflow. One of them is that the
dev_info() call will print the name of the device later, but AFAICT
we have only copied a pointer to the name earlier and the actual name
has been freed by the time it gets printed.
This removes the on-stack copy of the device and instead copies the
device name using kstrdup(). I'm ignoring the possible failure here
as both printk() and kfree() are able to deal with NULL pointers.
This fixes a bug in which the upper 32-bits of a 64-bit value which is
read by get_user() was lost on a 32-bit kernel.
While touching this code, split out pre-loading of %sr2 space register
and clean up code indent.
The ahash API modifies the request's callback function in order
to clean up after itself in some corner cases (unaligned final
and missing finup).
When the request is complete ahash will restore the original
callback and everything is fine. However, when the request gets
an EBUSY on a full queue, an EINPROGRESS callback is made while
the request is still ongoing.
In this case the ahash API will incorrectly call its own callback.
This patch fixes the problem by creating a temporary request
object on the stack which is used to relay EINPROGRESS back to
the original completion function.
This patch also adds code to preserve the original flags value.
The algif_aead completion function tries to deduce the aead_request
from the crypto_async_request argument. This is broken because
the API does not guarantee that the same request will be pased to
the completion function. Only the value of req->data can be used
in the completion function.
This patch fixes it by storing a pointer to sk in areq and using
that instead of passing in sk through req->data.
Fixes: 83094e5e9e49 ("crypto: af_alg - add async support to...") Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When function tracer has a pid filter, it adds a probe to sched_switch
to track if current task can be ignored. The probe checks the
ftrace_ignore_pid from current tr to filter tasks. But it misses to
delete the probe when removing an instance so that it can cause a crash
due to the invalid tr pointer (use-after-free).
The copy_page is optimized memcpy for page-alinged address. If it is
used with non-page aligned address, it can corrupt memory which means
system corruption. With zram, it can happen with
1. 64K architecture
2. partial IO
3. slub debug
Partial IO need to allocate a page and zram allocates it via kmalloc.
With slub debug, kmalloc(PAGE_SIZE) doesn't return page-size aligned
address. And finally, copy_page(mem, cmem) corrupts memory.
So, this patch changes it to memcpy.
Actuaully, we don't need to change zram_bvec_write part because zsmalloc
returns page-aligned address in case of PAGE_SIZE class but it's not
good to rely on the internal of zsmalloc.
Note:
When this patch is merged to stable, clear_page should be fixed, too.
Unfortunately, recent zram removes it by "same page merge" feature so
it's hard to backport this patch to -stable tree.
I will handle it when I receive the mail from stable tree maintainer to
merge this patch to backport.
Fixes: 42e99bd ("zram: optimize memory operations with clear_page()/copy_page()") Link: http://lkml.kernel.org/r/1492042622-12074-2-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It shouldn't have been included in a stable release.
Reported-by: Amit Pundir <amit.pundir@linaro.org> Cc: Felix Fietkau <nbd@nbd.name> Cc: John Crispin <john@phrozen.org> Cc: James Hogan <james.hogan@imgtec.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Without a bool string present, using "# CONFIG_DEVPORT is not set" in
defconfig files would not actually unset devport. This esnured that
/dev/port was always on, but there are reasons a user may wish to
disable it (smaller kernel, attack surface reduction) if it's not being
used. Adding a message here in order to make this user visible.
When two function probes are added to set_ftrace_filter, and then one of
them is removed, the update to the function locations is not performed, and
the record keeping of the function states are corrupted, and causes an
ftrace_bug() to occur.
This is easily reproducable by adding two probes, removing one, and then
adding it back again.
Fixes: 59df055f1991 ("ftrace: trace different functions with a different tracer") Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a report that after commit 27622b061eb4 ("cpufreq: Convert
to hotplug state machine"), the normal CPU offline/online cycle
fails on some platforms.
According to the ftrace result, this problem was triggered on
platforms using acpi-cpufreq as the default cpufreq driver,
and due to the lack of some ACPI freq method (eg. _PCT),
cpufreq_online() failed and returned a negative value, so the CPU
hotplug state machine rolled back the CPU online process. Actually,
from the user's perspective, the failure of cpufreq_online() should
not prevent that CPU from being brought up, although cpufreq might
not work on that CPU.
BTW, during system startup cpufreq_online() is not invoked via CPU
online but by the cpufreq device creation process, so the APs can be
brought up even though cpufreq_online() fails in that stage.
This patch ignores the return value of cpufreq_online/offline() and
lets the cpufreq framework deal with the failure. cpufreq_online()
itself will do a proper rollback in that case and if _PCT is missing,
the ACPI cpufreq driver will print a warning if the corresponding
debug options have been enabled.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=194581 Fixes: 27622b061eb4 ("cpufreq: Convert to hotplug state machine") Reported-and-tested-by: Tomasz Maciej Nowak <tmn505@gmail.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the PWM was not enabled at U-Boot loader, PWM could not work for
clock always disabled at PWM driver. The PWM clock is enabled at
beginning of pwm_apply(), but disabled at end of pwm_apply().
If the PWM was enabled at U-Boot loader, PWM clock is always enabled
unless closed by ATF. The pwm-backlight might turn off the power at
early suspend, should disable PWM clock for saving power consume.
It is important to provide opportunity to enable/disable clock at PWM
driver, the PWM consumer should ensure correct order to call PWM enable
and disable, and PWM driver ensure state of PWM clock synchronized with
PWM enabled state.
Fixes: 2bf1c98aa5a4 ("pwm: rockchip: Add support for atomic update") Signed-off-by: David Wu <david.wu@rock-chips.com> Reviewed-by: Boris Brezillon <boris.brezillon@free-electrons.com> Signed-off-by: Thierry Reding <thierry.reding@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Holding the reconfig_mutex over a potential userspace fault sets up a
lockdep dependency chain between filesystem-DAX and the libnvdimm ioctl
path. Move the user access outside of the lock.
[ INFO: possible circular locking dependency detected ]
4.11.0-rc3+ #13 Tainted: G W O
-------------------------------------------------------
fallocate/16656 is trying to acquire lock:
(&nvdimm_bus->reconfig_mutex){+.+.+.}, at: [<ffffffffa00080b1>] nvdimm_bus_lock+0x21/0x30 [libnvdimm]
but task is already holding lock:
(jbd2_handle){++++..}, at: [<ffffffff813b4944>] start_this_handle+0x104/0x460
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
Fixes: 62232e45f4a2 ("libnvdimm: control (ioctl) messages for nvdimm_bus and nvdimm devices") Cc: Dave Jiang <dave.jiang@intel.com> Reported-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a given blk-dpa allocation does not alias with any pmem ranges then
the full allocation should be accounted as busy space, not the size of
the current pmem contribution to the region.
The thinkos that led to this confusion was not realizing that the struct
resource management is already guaranteeing no collisions between pmem
allocations and blk allocations on the same dimm. Also, we do not try to
support blk allocations in aliased pmem holes.
This patch also fixes a case where the available blk goes negative.
Fixes: a1f3e4d6a0c3 ("libnvdimm, region: update nd_region_available_dpa() for multi-pmem support"). Reported-by: Dariusz Dokupil <dariusz.dokupil@intel.com> Reported-by: Dave Jiang <dave.jiang@intel.com> Reported-by: Vishal Verma <vishal.l.verma@intel.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Tested-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixes the mess observed in e.g. rsync over a noisy link we'd been
seeing since last Summer. What happens is that we copy part of
a datagram before noticing a checksum mismatch. Datagram will be
resent, all right, but we want the next try go into the same place,
not after it...
All this family of primitives (copy/checksum and copy a datagram
into destination) is "all or nothing" sort of interface - either
we get 0 (meaning that copy had been successful) or we get an
error (and no way to tell how much had been copied before we ran
into whatever error it had been). Make all of them leave iterator
unadvanced in case of errors - all callers must be able to cope
with that (an error might've been caught before the iterator had
been advanced), it costs very little to arrange, it's safer for
callers and actually fixes at least one bug in said callers.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Connecting to the backend isn't working reliably in xen-fbfront: in
case XenbusStateInitWait of the backend has been missed the backend
transition to XenbusStateConnected will trigger the connected state
only without doing the actions required when the backend has
connected.
This patch closes a race between se_lun deletion during configfs
unlink in target_fabric_port_unlink() -> core_dev_del_lun()
-> core_tpg_remove_lun(), when transport_clear_lun_ref() blocks
waiting for percpu_ref RCU grace period to finish, but a new
NodeACL mappedlun is added before the RCU grace period has
completed.
This can happen in target_fabric_mappedlun_link() because it
only checks for se_lun->lun_se_dev, which is not cleared until
after transport_clear_lun_ref() percpu_ref RCU grace period
finishes.
This bug originally manifested as NULL pointer dereference
OOPsen in target_stat_scsi_att_intr_port_show_attr_dev() on
v4.1.y code, because it dereferences lun->lun_se_dev without
a explicit NULL pointer check.
In post v4.1 code with target-core RCU conversion, the code
in target_stat_scsi_att_intr_port_show_attr_dev() no longer
uses se_lun->lun_se_dev, but the same race still exists.
To address the bug, go ahead and set se_lun>lun_shutdown as
early as possible in core_tpg_remove_lun(), and ensure new
NodeACL mappedlun creation in target_fabric_mappedlun_link()
fails during se_lun shutdown.
Reported-by: James Shen <jcs@datera.io> Cc: James Shen <jcs@datera.io> Tested-by: James Shen <jcs@datera.io> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We previously made sure that the reported disk capacity was less than
0xffffffff blocks when the kernel was not compiled with large sector_t
support (CONFIG_LBDAF). However, this check assumed that the capacity
was reported in units of 512 bytes.
Add a sanity check function to ensure that we only enable disks if the
entire reported capacity can be expressed in terms of sector_t.
Reported-by: Steve Magnani <steve.magnani@digidescorp.com> Cc: Bart Van Assche <Bart.VanAssche@sandisk.com> Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add fix to read correct register value for ISP82xx, during check for
register disconnect.ISP82xx has different base register.
Fixes: a465537ad1a4 ("qla2xxx: Disable the adapter and skip error recovery in case of register disconnect") Signed-off-by: Sawan Chandak <sawan.chandak@cavium.com> Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If device reports a small max_xfer_blocks and a zero opt_xfer_blocks, we
end up using BLK_DEF_MAX_SECTORS, which is wrong and r/w of that size
may get error.
[mkp: tweaked to avoid setting rw_max twice and added typecast]
Fixes: ca369d51b3e ("block/sd: Fix device-imposed transfer length limits") Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Kefeng Wang discovered that old versions of the QEMU CD driver would
return mangled mode data causing us to walk off the end of the buffer in
an attempt to parse it. Sanity check the returned mode sense data.
Reported-by: Kefeng Wang <wangkefeng.wang@huawei.com> Tested-by: Kefeng Wang <wangkefeng.wang@huawei.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Once upon a time back in 2009, a work-around was added to support
the GlobalSAN iSCSI initiator v3.3 for MacOSX, which during login
did not propose nor respond to MaxBurstLength, FirstBurstLength,
DefaultTime2Wait and DefaultTime2Retain keys.
The work-around in iscsi_check_proposer_for_optional_reply()
allowed the missing keys to be proposed, but did not require
waiting for a response before moving to full feature phase
operation. This allowed GlobalSAN v3.3 to work out-of-the
box, and for many years we didn't run into login interopt
issues with any other initiators..
Until recently, when Martin tried a QLogic 57840S iSCSI Offload
HBA on Windows 2016 which completed login, but subsequently
failed with:
Got unknown iSCSI OpCode: 0x43
The issue was QLogic MSFT side did not propose DefaultTime2Wait +
DefaultTime2Retain, so LIO proposes them itself, and immediately
transitions to full feature phase because of the GlobalSAN hack.
However, the QLogic MSFT side still attempts to respond to
DefaultTime2Retain + DefaultTime2Wait, even though LIO has set
ISCSI_FLAG_LOGIN_NEXT_STAGE3 + ISCSI_FLAG_LOGIN_TRANSIT
in last login response.
So while the QLogic MSFT side should have been proposing these
two keys to start, it was doing the correct thing per RFC-3720
attempting to respond to proposed keys before transitioning to
full feature phase.
All that said, recent versions of GlobalSAN iSCSI (v5.3.0.541)
does correctly propose the four keys during login, making the
original work-around moot.
So in order to allow QLogic MSFT to run unmodified as-is, go
ahead and drop this long standing work-around.
Reported-by: Martin Svec <martin.svec@zoner.cz> Cc: Martin Svec <martin.svec@zoner.cz> Cc: Himanshu Madhani <Himanshu.Madhani@cavium.com> Cc: Arun Easi <arun.easi@cavium.com> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch fixes a iscsi-target specific TMR reference leak
during session shutdown, that could occur when a TMR was
quiesced before the hand-off back to iscsi-target code
via transport_cmd_check_stop_to_fabric().
The reference leak happens because iscsit_free_cmd() was
incorrectly skipping the final target_put_sess_cmd() for
TMRs when transport_generic_free_cmd() returned zero because
the se_cmd->cmd_kref did not reach zero, due to the missing
se_cmd assignment in original code.
The result was iscsi_cmd and it's associated se_cmd memory
would be freed once se_sess->sess_cmd_map where released,
but the associated se_tmr_req was leaked and remained part
of se_device->dev_tmr_list.
This bug would manfiest itself as kernel paging request
OOPsen in core_tmr_lun_reset(), when a left-over se_tmr_req
attempted to dereference it's se_cmd pointer that had
already been released during normal session shutdown.
To address this bug, go ahead and treat ISCSI_OP_SCSI_CMD
and ISCSI_OP_SCSI_TMFUNC the same when there is an extra
se_cmd->cmd_kref to drop in iscsit_free_cmd(), and use
op_scsi to signal __iscsit_free_cmd() when the former
needs to clear any further iscsi related I/O state.
Reported-by: Rob Millner <rlm@daterainc.com> Cc: Rob Millner <rlm@daterainc.com> Reported-by: Chu Yuan Lin <cyl@datera.io> Cc: Chu Yuan Lin <cyl@datera.io> Tested-by: Chu Yuan Lin <cyl@datera.io> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On UEFI systems, the PCI subsystem is enumerated by the firmware,
and if a graphical framebuffer is exposed via a PCI device, its base
address and size are exposed to the OS via the Graphics Output
Protocol (GOP).
On arm64 PCI systems, the entire PCI hierarchy is reconfigured from
scratch at boot. This may result in the GOP framebuffer address to
become stale, if the BAR covering the framebuffer is modified. This
will cause the framebuffer to become unresponsive, and may in some
cases result in unpredictable behavior if the range is reassigned to
another device.
So add a non-x86 quirk to the EFI fb driver to find the BAR associated
with the GOP base address, and claim the BAR resource so that the PCI
core will not move it.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Peter Jones <pjones@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: leif.lindholm@linaro.org Cc: linux-efi@vger.kernel.org Cc: lorenzo.pieralisi@arm.com Fixes: 9822504c1fa5 ("efifb: Enable the efi-framebuffer platform driver ...") Link: http://lkml.kernel.org/r/20170404152744.26687-3-ard.biesheuvel@linaro.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The UEFI Specification permits Graphics Output Protocol (GOP) instances
without direct framebuffer access. This is indicated in the Mode structure
with a PixelFormat enumeration value of PIXEL_BLT_ONLY. Given that the
kernel does not know how to drive a Blt() only framebuffer (which is only
permitted before ExitBootServices() anyway), we should disregard such
framebuffers when looking for a GOP instance that is suitable for use as
the boot console.
So modify the EFI GOP initialization to not use a PIXEL_BLT_ONLY instance,
preventing attempts later in boot to use an invalid screen_info.lfb_base
address.
Signed-off-by: Eugene Cohen <eugene@hp.com>
[ Moved the Blt() only check into the loop and clarified that Blt() only GOPs are unusable by the kernel. ] Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: leif.lindholm@linaro.org Cc: linux-efi@vger.kernel.org Cc: lorenzo.pieralisi@arm.com Fixes: 9822504c1fa5 ("efifb: Enable the efi-framebuffer platform driver ...") Link: http://lkml.kernel.org/r/20170404152744.26687-2-ard.biesheuvel@linaro.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access
fault handling in pa_memcpy()") reimplements the pa_memcpy function.
Unfortunatelly, it makes the kernel unbootable. The crash happens in the
function ide_complete_cmd where memcpy is called with the same source
and destination address.
This patch fixes a few bugs in pa_memcpy:
* When jumping to .Lcopy_loop_16 for the first time, don't skip the
instruction "ldi 31,t0" (this bug made the kernel unbootable)
* Use the COND macro when comparing length, so that the comparison is
64-bit (a theoretical issue, in case the length is greater than
0xffffffff)
* Don't use the COND macro after the "extru" instruction (the PA-RISC
specification says that the upper 32-bits of extru result are undefined,
although they are set to zero in practice)
* Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault
* Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with
.Lcopy8_fault)
Commit 10c7e20b2ff3 (ACPI / scan: fix enumeration (visited) flags for
bus rescans) attempted to fix a problem with ACPI-based enumerateion
of I2C/SPI devices, but it forgot to ensure that the visited flag
will be set for all of the other enumerated devices, so fix that.
Fixes: 10c7e20b2ff3 (ACPI / scan: fix enumeration (visited) flags for bus rescans) Link: https://bugzilla.kernel.org/show_bug.cgi?id=194885 Reported-and-tested-by: Kevin Locke <kevin@kevinlocke.name> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While reviewing the -stable patch for commit 86ef58a4e35e "nfit,
libnvdimm: fix interleave set cookie calculation" Ben noted:
"This is returning an int, thus it's effectively doing a 32-bit
comparison and not the 64-bit comparison you say is needed."
Update the compare operation to be immune to this integer demotion problem.
Cc: Nicholas Moulin <nicholas.w.moulin@linux.intel.com> Fixes: 86ef58a4e35e ("nfit, libnvdimm: fix interleave set cookie calculation") Reported-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
vdso_enabled can be set to arbitrary integer values via the kernel command
line 'vdso32=' parameter or via 'sysctl abi.vsyscall32'.
load_vdso32() only maps VDSO if vdso_enabled == 1, but ARCH_DLINFO_IA32
merily checks for vdso_enabled != 0. As a consequence the AT_SYSINFO_EHDR
auxiliary vector for the VDSO_ENTRY is emitted with a NULL pointer which
causes a segfault when the application tries to use the VDSO.
Restrict the valid arguments on the command line and the sysctl to 0 and 1.
Before we rework the "pmem api" to stop abusing __copy_user_nocache()
for memcpy_to_pmem() we need to fix cases where we may strand dirty data
in the cpu cache. The problem occurs when copy_from_iter_pmem() is used
for arbitrary data transfers from userspace. There is no guarantee that
these transfers, performed by dax_iomap_actor(), will have aligned
destinations or aligned transfer lengths. Backstop the usage
__copy_user_nocache() with explicit cache management in these unaligned
cases.
Yes, copy_from_iter_pmem() is now too big for an inline, but addressing
that is saved for a later patch that moves the entirety of the "pmem
api" into the pmem driver directly.
Fixes: 5de490daec8b ("pmem: add copy_from_iter_pmem() and clear_pmem()") Cc: <x86@kernel.org> Cc: Jan Kara <jack@suse.cz> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Matthew Wilcox <mawilcox@microsoft.com> Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Toshi Kani <toshi.kani@hpe.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reserving a runtime region results in splitting the EFI memory
descriptors for the runtime region. This results in runtime region
descriptors with bogus memory mappings, leading to interesting crashes
like the following during a kexec:
This fixes Continuous Availability when errors during
file reopen are encountered.
cifs_user_readv and cifs_user_writev would wait for ever if
results of cifs_reopen_file are not stored and for later inspection.
In fact, results are checked and, in case of errors, a chain
of function calls leading to reads and writes to be scheduled in
a separate thread is skipped.
These threads will wake up the corresponding waiters once reads
and writes are done.
However, given the return value is not stored, when rc is checked
for errors a previous one (always zero) is inspected instead.
This leads to pending reads/writes added to the list, making
cifs_user_readv and cifs_user_writev wait for ever.
Signed-off-by: Germano Percossi <germano.percossi@citrix.com> Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com> Signed-off-by: Steve French <smfrench@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The NV4A (aka NV44A) is an oddity in the family. It only comes in AGP
and PCI varieties, rather than a core PCIE chip with a bridge for
AGP/PCI as necessary. As a result, it appears that the MMU is also
non-functional. For AGP cards, the vast majority of the NV4A lineup,
this worked out since we force AGP cards to use the nv04 mmu. However
for PCI variants, this did not work.
Switching to the NV04 MMU makes it work like a charm. Thanks to mwk for
the suggestion. This should be a no-op for NV4A AGP boards, as they were
using it already.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=70388 Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu> Signed-off-by: Ben Skeggs <bskeggs@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ 1337.483798] ================================================
[ 1337.483999] [ BUG: lock held when returning to user space! ]
[ 1337.484252] 4.11.0-rc6 #19 Not tainted
[ 1337.484423] ------------------------------------------------
[ 1337.484626] mount/14766 is leaving the kernel with locks still held!
[ 1337.484841] 1 lock held by mount/14766:
[ 1337.485017] #0: (&type->s_umount_key#33/1){+.+.+.}, at: [<ffffffff8124171f>] sget_userns+0x2af/0x520
Caught by xfstests generic/413 which tried to mount with the unsupported
mount option dax. Then xfstests generic/422 ran sync which deadlocks.
Signed-off-by: Martin Brandenburg <martin@omnibond.com> Acked-by: Mike Marshall <hubcap@omnibond.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Now 64K page system, zsamlloc has 257 classes so 8 class bit is not
enough. With that, it corrupts the system when zsmalloc stores
65536byte data(ie, index number 256) so that this patch increases class
bit for simple fix for stable backport. We should clean up this mess
soon.
Both MADV_DONTNEED and MADV_FREE handled with down_read(mmap_sem).
It's critical to not clear pmd intermittently while handling MADV_FREE
to avoid race with MADV_DONTNEED:
CPU0: CPU1:
madvise_free_huge_pmd()
pmdp_huge_get_and_clear_full()
madvise_dontneed()
zap_pmd_range()
pmd_trans_huge(*pmd) == 0 (without ptl)
// skip the pmd
set_pmd_at();
// pmd is re-established
It results in MADV_DONTNEED skipping the pmd, leaving it not cleared.
It violates MADV_DONTNEED interface and can result is userspace
misbehaviour.
Basically it's the same race as with numa balancing in
change_huge_pmd(), but a bit simpler to mitigate: we don't need to
preserve dirty/young flags here due to MADV_FREE functionality.
[kirill.shutemov@linux.intel.com: Urgh... Power is special again] Link: http://lkml.kernel.org/r/20170303102636.bhd2zhtpds4mt62a@black.fi.intel.com Link: http://lkml.kernel.org/r/20170302151034.27829-4-kirill.shutemov@linux.intel.com Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Hillf Danton <hillf.zj@alibaba-inc.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For the bidirectional case, the Data-Out buffer blocks will always at
the head of the tcmu_cmd's bitmap, and before gathering the Data-In
buffer, first of all it should skip the Data-Out ones, or the device
supporting BIDI commands won't work.
The t_data_nents and t_bidi_data_nents are the numbers of the
segments, but it couldn't be sure the block size equals to size
of the segment.
For the worst case, all the blocks are discontiguous and there
will need the same number of iovecs, that's to say: blocks == iovs.
So here just set the number of iovs to block count needed by tcmu
cmd.
Tested-by: Ilias Tsitsimpis <iliastsi@arrikto.com> Reviewed-by: Mike Christie <mchristi@redhat.com> Signed-off-by: Xiubo Li <lixiubo@cmss.chinamobile.com> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If there has BIDI data, its first iov[] will overwrite the last
iov[] for se_cmd->t_data_sg.
To fix this, we can just increase the iov pointer, but this may
introuduce a new memory leakage bug: If the se_cmd->data_length
and se_cmd->t_bidi_data_sg->length are all not aligned up to the
DATA_BLOCK_SIZE, the actual length needed maybe larger than just
sum of them.
So, this could be avoided by rounding all the data lengthes up
to DATA_BLOCK_SIZE.
Reviewed-by: Mike Christie <mchristi@redhat.com> Tested-by: Ilias Tsitsimpis <iliastsi@arrikto.com> Reviewed-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com> Signed-off-by: Xiubo Li <lixiubo@cmss.chinamobile.com> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Creation of a kthread goes through a couple interlocked stages between
the kthread itself and its creator. Once the new kthread starts
running, it initializes itself and wakes up the creator. The creator
then can further configure the kthread and then let it start doing its
job by waking it up.
In this configuration-by-creator stage, the creator is the only one
that can wake it up but the kthread is visible to userland. When
altering the kthread's attributes from userland is allowed, this is
fine; however, for cases where CPU affinity is critical,
kthread_bind() is used to first disable affinity changes from userland
and then set the affinity. This also prevents the kthread from being
migrated into non-root cgroups as that can affect the CPU affinity and
many other things.
Unfortunately, the cgroup side of protection is racy. While the
PF_NO_SETAFFINITY flag prevents further migrations, userland can win
the race before the creator sets the flag with kthread_bind() and put
the kthread in a non-root cgroup, which can lead to all sorts of
problems including incorrect CPU affinity and starvation.
This bug got triggered by userland which periodically tries to migrate
all processes in the root cpuset cgroup to a non-root one. Per-cpu
workqueue workers got caught while being created and ended up with
incorrected CPU affinity breaking concurrency management and sometimes
stalling workqueue execution.
This patch adds task->no_cgroup_migration which disallows the task to
be migrated by userland. kthreadd starts with the flag set making
every child kthread start in the root cgroup with migration
disallowed. The flag is cleared after the kthread finishes
initialization by which time PF_NO_SETAFFINITY is set if the kthread
should stay in the root cgroup.
It'd be better to wait for the initialization instead of failing but I
couldn't think of a way of implementing that without adding either a
new PF flag, or sleeping and retrying from waiting side. Even if
userland depends on changing cgroup membership of a kthread, it either
has to be synchronized with kthread_create() or periodically repeat,
so it's unlikely that this would break anything.
v2: Switch to a simpler implementation using a new task_struct bit
field suggested by Oleg.
Signed-off-by: Tejun Heo <tj@kernel.org> Suggested-by: Oleg Nesterov <oleg@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Reported-and-debugged-by: Chris Mason <clm@fb.com> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Boris Brezillon [Thu, 2 Feb 2017 10:37:36 +0000 (11:37 +0100)]
ARM: dts: bcm283x: Add HDMI audio related properties
Add the dmas and dma-names properties to support HDMI audio.
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com> Signed-off-by: Eric Anholt <eric@anholt.net>
(cherry picked from commit d46d2c6380c10e80e99f6af9067356128bffac6b)
Eric Anholt [Mon, 27 Feb 2017 20:28:02 +0000 (12:28 -0800)]
drm/vc4: Add HDMI audio support
The HDMI encoder IP embeds all needed blocks to output audio, with a
custom DAI called MAI moving audio between the two parts of the HDMI
core. This driver now exposes a sound card to let users stream audio
to their display.
Using the hdmi-codec driver has been considered here, but MAI meant
having to significantly rework hdmi-codec, and it would have left
little shared code with the I2S mode anyway.
The encoder requires that the audio be SPDIF-formatted frames only,
which alsalib will format-convert for us.
This patch is the combined work of Eric Anholt (initial register setup
with a separate dmaengine driver and using simple-audio-card) and
Boris Brezillon (moving it all into HDMI, massive debug to get it
actually working), and which Eric has the permission to release.
v2: Drop "-audio" from sound card name, since that's already implied
(suggestion by Boris)
Eric Anholt [Tue, 7 Mar 2017 22:51:03 +0000 (14:51 -0800)]
drm/vc4: Allow using more than 256MB of CMA memory.
Until now, we've had to limit Raspberry Pi to 256MB of CMA memory to
keep from triggering the hardware addressing bug between of the tile
binner of the tile alloc memory (where the top 4 bits come from the
tile state data array's address).
To work around that and allow more memory to be reserved for graphics,
allocate a single BO to store tile state data arrays and tile
alloc/overflow memory while the GPU is active, and make sure that that
one BO doesn't happen to cross a 256MB boundary. With that in place,
we can allocate textures and shaders anywhere in system memory (still
contiguous, of course).
Add compat ioctl support to dma-buf. This lets one to use DMA_BUF_IOCTL_SYNC
ioctl from 32bit application on 64bit kernel. Data structures for both 32
and 64bit modes are same, so there is no need for additional translation
layer.
Save the qp context flags byte containing the flag disabling vlan stripping
in the RESET to INIT qp transition, rather than in the INIT to RTR
transition. Per the firmware spec, the flags in this byte are active
in the RESET to INIT transition.
As a result of saving the flags in the incorrect qp transition, when
switching dynamically from VGT to VST and back to VGT, the vlan
remained stripped (as is required for VST) and did not return to
not-stripped (as is required for VGT).
Fixes: f0f829bf42cd ("net/mlx4_core: Add immediate activate for VGT->VST->VGT") Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Tariq Toukan <tariqt@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In function mlx4_cq_completion() and mlx4_cq_event(), the
radix_tree_lookup requires a rcu_read_lock.
This is mandatory: if another core frees the CQ, it could
run the radix_tree_node_rcu_free() call_rcu() callback while
its being used by the radix tree lookup function.
Additionally, in function mlx4_cq_event(), since we are adding
the rcu lock around the radix-tree lookup, we no longer need to take
the spinlock. Also, the synchronize_irq() call for the async event
eliminates the need for incrementing the cq reference count in
mlx4_cq_event().
Other changes:
1. In function mlx4_cq_free(), replace spin_lock_irq with spin_lock:
we no longer take this spinlock in the interrupt context.
The spinlock here, therefore, simply protects against different
threads simultaneously invoking mlx4_cq_free() for different cq's.
2. In function mlx4_cq_free(), we move the radix tree delete to before
the synchronize_irq() calls. This guarantees that we will not
access this cq during any subsequent interrupts, and therefore can
safely free the CQ after the synchronize_irq calls. The rcu_read_lock
in the interrupt handlers only needs to protect against corrupting the
radix tree; the interrupt handlers may access the cq outside the
rcu_read_lock due to the synchronize_irq calls which protect against
premature freeing of the cq.
3. In function mlx4_cq_event(), we change the mlx_warn message to mlx4_dbg.
4. We leave the cq reference count mechanism in place, because it is
still needed for the cq completion tasklet mechanism.
Fixes: 6d90aa5cf17b ("net/mlx4_core: Make sure there are no pending async events when freeing CQ") Fixes: 225c7b1feef1 ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters") Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Matan Barak <matanb@mellanox.com> Signed-off-by: Tariq Toukan <tariqt@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On a system with a defective USB device connected to an USB hub,
an endless sequence of port connect events was observed. The sequence
of events as observed is as follows:
- Port reports connected event (port status=USB_PORT_STAT_CONNECTION).
- Event handler debounces port and resets it by calling hub_port_reset().
- hub_port_reset() calls hub_port_wait_reset() to wait for the reset
to complete.
- The reset completes, but USB_PORT_STAT_CONNECTION is not immediately
set in the port status register.
- hub_port_wait_reset() returns -ENOTCONN.
- Port initialization sequence is aborted.
- A few milliseconds later, the port again reports a connected event,
and the sequence repeats.
This continues either forever or, randomly, stops if the connection
is already re-established when the port status is read. It results in
a high rate of udev events. This in turn destabilizes userspace since
the above sequence holds the device mutex pretty much continuously
and prevents userspace from actually reading the device status.
To prevent the problem from happening, let's wait for the connection
to be re-established after a port reset. If the device was actually
disconnected, the code will still return an error, but it will do so
only after the long reset timeout.
Cc: Douglas Anderson <dianders@chromium.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While stressing memory and IO at the same time we changed SMT settings,
we were able to consistently trigger deadlocks in the mm system, which
froze the entire machine.
I think that under memory stress conditions, the large allocations
performed by blk_mq_init_rq_map may trigger a reclaim, which stalls
waiting on the block layer remmaping completion, thus deadlocking the
system. The trace below was collected after the machine stalled,
waiting for the hotplug event completion.
The simplest fix for this is to make allocations in this path
non-reclaimable, with GFP_NOIO. With this patch, We couldn't hit the
issue anymore.
This should apply on top of Jens's for-next branch cleanly.
Changes since v1:
- Use GFP_NOIO instead of GFP_NOWAIT.
Subtracting tp_sizeof_priv from tp_block_size and casting to int
to check whether one is less then the other doesn't always work
(both of them are unsigned ints).
Compare them as is instead.
Also cast tp_sizeof_priv to u64 before using BLK_PLUS_PRIV, as
it can overflow inside BLK_PLUS_PRIV otherwise.
Signed-off-by: Andrey Konovalov <andreyknvl@google.com> Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jani Nikula <jani.nikula@linux.intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Eric Blau <eblau1@gmail.com> Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
projects / linux-4.9.git / log