Since commit 71ce391dfb784 ("net: mvpp2: enable proper per-CPU TX
buffers unmapping"), we are not correctly DMA unmapping TX buffers for
fragments.
Indeed, the mvpp2_txq_inc_put() function only stores in the
txq_cpu->tx_buffs[] array the physical address of the buffer to be
DMA-unmapped when skb != NULL. In addition, when DMA-unmapping, we use
skb_headlen(skb) to get the size to be unmapped. Both of this works fine
for TX descriptors that are associated directly to a SKB, but not the
ones that are used for fragments, with a NULL pointer as skb:
- We have a NULL physical address when calling DMA unmap
- skb_headlen(skb) crashes because skb is NULL
This causes random crashes when fragments are used.
To solve this problem, we need to:
- Store the physical address of the buffer to be unmapped
unconditionally, regardless of whether it is tied to a SKB or not.
- Store the length of the buffer to be unmapped, which requires a new
field.
Instead of adding a third array to store the length of the buffer to be
unmapped, and as suggested by David Miller, this commit refactors the
tx_buffs[] and tx_skb[] arrays of 'struct mvpp2_txq_pcpu' into a
separate structure 'mvpp2_txq_pcpu_buf', to which a 'size' field is
added. Therefore, instead of having three arrays to allocate/free, we
have a single one, which also improve data locality, reducing the
impact on the CPU cache.
Fixes: 71ce391dfb784 ("net: mvpp2: enable proper per-CPU TX buffers unmapping") Reported-by: Raphael G <raphael.glon@corp.ovh.com> Cc: Raphael G <raphael.glon@corp.ovh.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Both damn things interpret userland pointers embedded into the payload;
worse, they are actually traversing those. Leaving aside the bad
API design, this is very much _not_ safe to call with KERNEL_DS.
Bail out early if that happens.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently it is impossible to edit the value of a config symbol with a
prompt longer than (terminal width - 2) characters. dialog_inputbox()
calculates a negative x-offset for the input window and newwin() fails
as this is invalid. It also doesn't check for this failure, so it
busy-loops calling wgetch(NULL) which immediately returns -1.
The additions in the offset calculations also don't match the intended
size of the window.
Limit the window size and calculate the offset similarly to
show_scroll_win().
Fixes: 692d97c380c6 ("kconfig: new configuration interface (nconfig)") Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Don't free the cmd in tcmu_check_expired_cmd, it's still referenced by
an entry in our cmd_id->cmd idr. If userspace ever resumes processing,
tcmu_handle_completions() will use the now-invalid cmd pointer.
Instead, don't free cmd. It will be freed by tcmu_handle_completion() if
userspace ever recovers, or tcmu_free_device if not.
Reported-by: Bryant G Ly <bgly@us.ibm.com> Tested-by: Bryant G Ly <bgly@us.ibm.com> Signed-off-by: Andy Grover <agrover@redhat.com> Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
PowerPC's "cmp" instruction has four operands. Normally people write
"cmpw" or "cmpd" for the second cmp operand 0 or 1. But, frequently
people forget, and write "cmp" with just three operands.
With older binutils this is silently accepted as if this was "cmpw",
while often "cmpd" is wanted. With newer binutils GAS will complain
about this for 64-bit code. For 32-bit code it still silently assumes
"cmpw" is what is meant.
In this instance the code comes directly from ISA v2.07, including the
cmp, but cmpd is correct. Backport to stable so that new toolchains can
build old kernels.
Fixes: 948cf67c4726 ("powerpc: Add NAP mode support on Power7 in HV mode") Reviewed-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> Signed-off-by: Segher Boessenkool <segher@kernel.crashing.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Joel Stanley <joel@jms.id.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
GCC 5 generates different code for this bootwrapper null check that
causes the PS3 to hang very early in its bootup. This check is of
limited value, so just get rid of it.
What matters when deciding if we should make a page uptodate is
not how much we _wanted_ to copy, but how much we actually have
copied. As it is, on architectures that do not zero tail on
short copy we can leave uninitialized data in page marked uptodate.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After sending an authorizer (ceph_x_authorize_a + ceph_x_authorize_b),
the client gets back a ceph_x_authorize_reply, which it is supposed to
verify to ensure the authenticity and protect against replay attacks.
The code for doing this is there (ceph_x_verify_authorizer_reply(),
ceph_auth_verify_authorizer_reply() + plumbing), but it is never
invoked by the the messenger.
AFAICT this goes back to 2009, when ceph authentication protocols
support was added to the kernel client in 4e7a5dcd1bba ("ceph:
negotiate authentication protocol; implement AUTH_NONE protocol").
The second param of ceph_connection_operations::verify_authorizer_reply
is unused all the way down. Pass 0 to facilitate backporting, and kill
it in the next commit.
One some systems, the firmware does not allow certain PCI devices to be put
in deep D-states. This can cause problems for wakeup signalling, if the
device does not support PME# in the deepest allowed suspend state. For
example, Pierre reports that on his system, ACPI does not permit his xHCI
host controller to go into D3 during runtime suspend -- but D3 is the only
state in which the controller can generate PME# signals. As a result, the
controller goes into runtime suspend but never wakes up, so it doesn't work
properly. USB devices plugged into the controller are never detected.
If the device relies on PME# for wakeup signals but is not capable of
generating PME# in the target state, the PCI core should accurately report
that it cannot do wakeup from runtime suspend. This patch modifies the
pci_dev_run_wake() routine to add this check.
Reported-by: Pierre de Villemereuil <flyos@mailoo.org> Tested-by: Pierre de Villemereuil <flyos@mailoo.org> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> CC: Lukas Wunner <lukas@wunner.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We were assigning I2C bus controller instead of client as parent device.
Besides being logically wrong, it messed up with devm handling of input
device. As a result we were leaving input device and event node behind
after rmmod-ing the driver, which lead to a kernel oops if one were to
access the event node later.
Let's remove the assignment and rely on devm_input_allocate_device() to
set it up properly for us.
This fixes a lockup at device probing which happens on some solo6010
hardware samples. This is a regression introduced by commit e1ceb25a1569
("[media] SOLO6x10: remove unneeded register locking and barriers")
The observed lockup happens in solo_set_motion_threshold() called from
solo_motion_config().
This extra "flushing" is not fundamentally needed for every write, but
apparently the code in driver assumes such behaviour at last in some
places.
Code that dereferences the struct net_device ip_ptr member must be
protected with an in_dev_get() / in_dev_put() pair. Hence insert
calls to these functions.
Fixes: commit 7b85627b9f02 ("IB/cma: IBoE (RoCE) IP-based GID addressing") Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Reviewed-by: Moni Shoua <monis@mellanox.com> Cc: Or Gerlitz <ogerlitz@mellanox.com> Cc: Roland Dreier <roland@purestorage.com> Signed-off-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The array ib_mad_mgmt_class_table.method_table has MAX_MGMT_CLASS
(80) elements. Hence compare the array index with that value instead
of with IB_MGMT_MAX_METHODS (128). This patch avoids that Coverity
reports the following:
Overrunning array class->method_table of 80 8-byte elements at element index 127 (byte offset 1016) using index convert_mgmt_class(mad_hdr->mgmt_class) (which evaluates to 127).
Fixes: commit b7ab0b19a85f ("IB/mad: Verify mgmt class in received MADs") Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Cc: Sean Hefty <sean.hefty@intel.com> Reviewed-by: Hal Rosenstock <hal@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Both the wakeup and irqsoff tracers can use the function graph tracer when
the display-graph option is set. The problem is that they ignore the notrace
file, and record the entry of functions that would be ignored by the
function_graph tracer. This causes the trace->depth to be recorded into the
ring buffer. The set_graph_notrace uses a trick by adding a large negative
number to the trace->depth when a graph function is to be ignored.
On trace output, the graph function uses the depth to record a stack of
functions. But since the depth is negative, it accesses the array with a
negative number and causes an out of bounds access that can cause a kernel
oops or corrupt data.
Have the print functions handle cases where a tracer still records functions
even when they are in set_graph_notrace.
Also add warnings if the depth is below zero before accessing the array.
Note, the function graph logic will still prevent the return of these
functions from being recorded, which means that they will be left hanging
without a return. For example:
Signed-off-by: Marcos Paulo de Souza <marcos.souza.org@gmail.com> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With new binutils, gcc may get smart with its optimization and change a jmp
from a 5 byte jump to a 2 byte one even though it was jumping to a global
function. But that global function existed within a 2 byte radius, and gcc
was able to optimize it. Unfortunately, that jump was also being modified
when function graph tracing begins. Since ftrace expected that jump to be 5
bytes, but it was only two, it overwrote code after the jump, causing a
crash.
This was fixed for x86_64 with commit 8329e818f149, with the same subject as
this commit, but nothing was done for x86_32.
Fixes: d61f82d06672 ("ftrace: use dynamic patching for updating mcount calls") Reported-by: Colin Ian King <colin.king@canonical.com> Tested-by: Colin Ian King <colin.king@canonical.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When L2 exits to L0 due to "exception or NMI", software exceptions
(#BP and #OF) for which L1 has requested an intercept should be
handled by L1 rather than L0. Previously, only hardware exceptions
were forwarded to L1.
Signed-off-by: Jim Mattson <jmattson@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The hashed page table MMU in POWER processors can update the R
(reference) and C (change) bits in a HPTE at any time until the
HPTE has been invalidated and the TLB invalidation sequence has
completed. In kvmppc_h_protect, which implements the H_PROTECT
hypercall, we read the HPTE, modify the second doubleword,
invalidate the HPTE in memory, do the TLB invalidation sequence,
and then write the modified value of the second doubleword back
to memory. In doing so we could overwrite an R/C bit update done
by hardware between when we read the HPTE and when the TLB
invalidation completed. To fix this we re-read the second
doubleword after the TLB invalidation and OR in the (possibly)
new values of R and C. We can use an OR since hardware only ever
sets R and C, never clears them.
This race was found by code inspection. In principle this bug could
cause occasional guest memory corruption under host memory pressure.
Fixes: a8606e20e41a ("KVM: PPC: Handle some PAPR hcalls in the kernel", 2011-06-29) Signed-off-by: Paul Mackerras <paulus@ozlabs.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When switching from/to a guest that has a transaction in progress,
we need to save/restore the checkpointed register state. Although
XER is part of the CPU state that gets checkpointed, the code that
does this saving and restoring doesn't save/restore XER.
This fixes it by saving and restoring the XER. To allow userspace
to read/write the checkpointed XER value, we also add a new ONE_REG
specifier.
The visible effect of this bug is that the guest may see its XER
value being corrupted when it uses transactions.
Fixes: e4e38121507a ("KVM: PPC: Book3S HV: Add transactional memory support") Fixes: 0a8eccefcb34 ("KVM: PPC: Book3S HV: Add missing code for transaction reclaim on guest exit") Signed-off-by: Paul Mackerras <paulus@ozlabs.org> Reviewed-by: Thomas Huth <thuth@redhat.com> Signed-off-by: Paul Mackerras <paulus@ozlabs.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Current implementation employ 16bit counter of active stripes in lower
bits of bio->bi_phys_segments. If request is big enough to overflow
this counter bio will be completed and freed too early.
Fortunately this not happens in default configuration because several
other limits prevent that: stripe_cache_size * nr_disks effectively
limits count of active stripes. And small max_sectors_kb at lower
disks prevent that during normal read/write operations.
Overflow easily happens in discard if it's enabled by module parameter
"devices_handle_discard_safely" and stripe_cache_size is set big enough.
This patch limits requests size with 256Mb - 8Kb to prevent overflows.
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru> Cc: Shaohua Li <shli@kernel.org> Cc: Neil Brown <neilb@suse.com> Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The use of IRQF_ONESHOT when registering an interrupt handler with
request_irq() is non-sensical.
Not only that, it also prevents the handler from being threaded when it
otherwise should be w/ IRQ_FORCED_THREADING is enabled. This causes the
following deadlock observed by Sean Nyekjaer on -rt:
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP ARM
[..]
rt_spin_lock_slowlock from queue_kthread_work
queue_kthread_work from sc16is7xx_irq
sc16is7xx_irq [sc16is7xx] from handle_irq_event_percpu
handle_irq_event_percpu from handle_irq_event
handle_irq_event from handle_level_irq
handle_level_irq from generic_handle_irq
generic_handle_irq from mxc_gpio_irq_handler
mxc_gpio_irq_handler from mx3_gpio_irq_handler
mx3_gpio_irq_handler from generic_handle_irq
generic_handle_irq from __handle_domain_irq
__handle_domain_irq from gic_handle_irq
gic_handle_irq from __irq_svc
__irq_svc from rt_spin_unlock
rt_spin_unlock from kthread_worker_fn
kthread_worker_fn from kthread
kthread from ret_from_fork
Fixes: 9e6f4ca3e567 ("sc16is7xx: use kthread_worker for tx_work and irq") Reported-by: Sean Nyekjaer <sean.nyekjaer@prevas.dk> Signed-off-by: Josh Cartwright <joshc@ni.com> Cc: linux-rt-users@vger.kernel.org Cc: Jakub Kicinski <moorray3@wp.pl> Cc: linux-serial@vger.kernel.org Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Julia Cartwright <julia@ni.com> Acked-by: Jakub Kicinski <kubakici@wp.pl> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The buffer for iucv_message_receive() needs to be below 2 GB. In
__iucv_message_receive(), the buffer address is casted to an u32, which
would result in either memory corruption or an addressing exception when
using addresses >= 2 GB.
Fix this by using GFP_DMA for the buffer allocation.
When you use the firmware usermode helper fallback with a timeout value set to a
value greater than INT_MAX (2147483647) a cast overflow issue causes the
timeout value to go negative and breaks all usermode helper loading. This
regression was introduced through commit 68ff2a00dbf5 ("firmware_loader:
handle timeout via wait_for_completion_interruptible_timeout()") on kernel
v4.0.
The firmware_class drivers relies on the firmware usermode helper
fallback as a mechanism to look for firmware if the direct filesystem
search failed only if:
a) You've enabled CONFIG_FW_LOADER_USER_HELPER_FALLBACK (not many distros):
Then all of these callers will rely on the fallback mechanism in case
the firmware is not found through an initial direct filesystem lookup:
o request_firmware()
o request_firmware_into_buf()
o request_firmware_nowait()
b) If you've only enabled CONFIG_FW_LOADER_USER_HELPER (most distros):
Then only callers using request_firmware_nowait() with the second
argument set to false, this explicitly is requesting the UMH firmware
fallback to be relied on in case the first filesystem lookup fails.
Using Coccinelle SmPL grammar we have identified only two drivers
explicitly requesting the UMH firmware fallback mechanism:
Since most distributions only enable CONFIG_FW_LOADER_USER_HELPER the
biggest impact of this regression are users of the dell_rbu and
leds-lp55xx-common device driver which required the UMH to find their
respective needed firmwares.
The default timeout for the UMH is set to 60 seconds always, as of
commit 68ff2a00dbf5 ("firmware_loader: handle timeout via
wait_for_completion_interruptible_timeout()") the timeout was bumped
to MAX_JIFFY_OFFSET ((LONG_MAX >> 1)-1). Additionally the MAX_JIFFY_OFFSET
value was also used if the timeout was configured by a user to 0.
A max value of INT_MAX (2147483647) seconds is therefore implicit due to the
another cast with simple_strtol().
This fixes the secondary cast (the first one is simple_strtol() but its an
issue only by forcing an implicit limit) by re-using the timeout variable and
only setting retval in appropriate cases.
Lastly worth noting systemd had ripped out the UMH firmware fallback
mechanism from udev since udev 2014 via commit be2ea723b1d023b3d
("udev: remove userspace firmware loading support"), so as of systemd v217.
Signed-off-by: Yves-Alexis Perez <corsac@corsac.net> Fixes: 68ff2a00dbf5 "firmware_loader: handle timeout via wait_for_completion_interruptible_timeout()" Cc: Luis R. Rodriguez <mcgrof@kernel.org> Cc: Ming Lei <ming.lei@canonical.com> Cc: Bjorn Andersson <bjorn.andersson@linaro.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Acked-by: Luis R. Rodriguez <mcgrof@kernel.org> Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
[mcgrof@kernel.org: gave commit log a whole lot of love] Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
An ARC700 customer reported linux boot crashes when upgrading to bigger
L1 dcache (64K from 32K). Turns out they had an aliasing VIPT config and
current code only assumed 2 colours, while theirs had 4. So default to 4
colours and complain if there are fewer. Ideally this needs to be a
Kconfig option, but heck that's too much of hassle for a single user.
A race between scanning and fc_remote_port_delete() may result in a
permanent stop if the device gets blocked before scsi_sysfs_add_sdev()
and unblocked after. The reason is that blocking a device sets both the
SDEV_BLOCKED state and the QUEUE_FLAG_STOPPED. However,
scsi_sysfs_add_sdev() unconditionally sets SDEV_RUNNING which causes the
device to be ignored by scsi_target_unblock() and thus never have its
QUEUE_FLAG_STOPPED cleared leading to a device which is apparently
running but has a stopped queue.
We actually have two places where SDEV_RUNNING is set: once in
scsi_add_lun() which respects the blocked flag and once in
scsi_sysfs_add_sdev() which doesn't. Since the second set is entirely
spurious, simply remove it to fix the problem.
Reported-by: Zengxi Chen <chenzengxi@huawei.com> Signed-off-by: Wei Fang <fangwei1@huawei.com> Reviewed-by: Ewan D. Milne <emilne@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It is unavoidable that zfcp_scsi_queuecommand() has to finish requests
with DID_IMM_RETRY (like fc_remote_port_chkready()) during the time
window when zfcp detected an unavailable rport but
fc_remote_port_delete(), which is asynchronous via
zfcp_scsi_schedule_rport_block(), has not yet blocked the rport.
However, for the case when the rport becomes available again, we should
prevent unblocking the rport too early. In contrast to other FCP LLDDs,
zfcp has to open each LUN with the FCP channel hardware before it can
send I/O to a LUN. So if a port already has LUNs attached and we
unblock the rport just after port recovery, recoveries of LUNs behind
this port can still be pending which in turn force
zfcp_scsi_queuecommand() to unnecessarily finish requests with
DID_IMM_RETRY.
This also opens a time window with unblocked rport (until the followup
LUN reopen recovery has finished). If a scsi_cmnd timeout occurs during
this time window fc_timed_out() cannot work as desired and such command
would indeed time out and trigger scsi_eh. This prevents a clean and
timely path failover. This should not happen if the path issue can be
recovered on FC transport layer such as path issues involving RSCNs.
Fix this by only calling zfcp_scsi_schedule_rport_register(), to
asynchronously trigger fc_remote_port_add(), after all LUN recoveries as
children of the rport have finished and no new recoveries of equal or
higher order were triggered meanwhile. Finished intentionally includes
any recovery result no matter if successful or failed (still unblock
rport so other successful LUNs work). For simplicity, we check after
each finished LUN recovery if there is another LUN recovery pending on
the same port and then do nothing. We handle the special case of a
successful recovery of a port without LUN children the same way without
changing this case's semantics.
For debugging we introduce 2 new trace records written if the rport
unblock attempt was aborted due to still unfinished or freshly triggered
recovery. The records are only written above the default trace level.
Benjamin noticed the important special case of new recovery that can be
triggered between having given up the erp_lock and before calling
zfcp_erp_action_cleanup() within zfcp_erp_strategy(). We must avoid the
following sequence:
ERP thread rport_work other context
------------------------- -------------- --------------------------------
port is unblocked, rport still blocked,
due to pending/running ERP action,
so ((port->status & ...UNBLOCK) != 0)
and (port->rport == NULL)
unlock ERP
zfcp_erp_action_cleanup()
case ZFCP_ERP_ACTION_REOPEN_LUN:
zfcp_erp_try_rport_unblock()
((status & ...UNBLOCK) != 0) [OLD!]
zfcp_erp_port_reopen()
lock ERP
zfcp_erp_port_block()
port->status clear ...UNBLOCK
unlock ERP
zfcp_scsi_schedule_rport_block()
port->rport_task = RPORT_DEL
queue_work(rport_work)
zfcp_scsi_rport_work()
(port->rport_task != RPORT_ADD)
port->rport_task = RPORT_NONE
zfcp_scsi_rport_block()
if (!port->rport) return
zfcp_scsi_schedule_rport_register()
port->rport_task = RPORT_ADD
queue_work(rport_work)
zfcp_scsi_rport_work()
(port->rport_task == RPORT_ADD)
port->rport_task = RPORT_NONE
zfcp_scsi_rport_register()
(port->rport == NULL)
rport = fc_remote_port_add()
port->rport = rport;
Now the rport was erroneously unblocked while the zfcp_port is blocked.
This is another situation we want to avoid due to scsi_eh
potential. This state would at least remain until the new recovery from
the other context finished successfully, or potentially forever if it
failed. In order to close this race, we take the erp_lock inside
zfcp_erp_try_rport_unblock() when checking the status of zfcp_port or
LUN. With that, the possible corresponding rport state sequences would
be: (unblock[ERP thread],block[other context]) if the ERP thread gets
erp_lock first and still sees ((port->status & ...UNBLOCK) != 0),
(block[other context],NOP[ERP thread]) if the ERP thread gets erp_lock
after the other context has already cleard ...UNBLOCK from port->status.
Since checking fields of struct erp_action is unsafe because they could
have been overwritten (re-used for new recovery) meanwhile, we only
check status of zfcp_port and LUN since these are only changed under
erp_lock elsewhere. Regarding the check of the proper status flags (port
or port_forced are similar to the shown adapter recovery):
Hence, we should check for both UNBLOCK and ERP_INUSE because they are
interleaved. Also we need to explicitly check ERP_FAILED for the link
down case which currently does not clear the UNBLOCK flag in
zfcp_fsf_link_down_info_eval().
Signed-off-by: Steffen Maier <maier@linux.vnet.ibm.com> Fixes: 8830271c4819 ("[SCSI] zfcp: Dont fail SCSI commands when transitioning to blocked fc_rport") Fixes: a2fa0aede07c ("[SCSI] zfcp: Block FC transport rports early on errors") Fixes: 5f852be9e11d ("[SCSI] zfcp: Fix deadlock between zfcp ERP and SCSI") Fixes: 338151e06608 ("[SCSI] zfcp: make use of fc_remote_port_delete when target port is unavailable") Fixes: 3859f6a248cb ("[PATCH] zfcp: add rports to enable scsi_add_device to work again") Reviewed-by: Benjamin Block <bblock@linux.vnet.ibm.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since quite a while, Linux issues enough SCSI commands per scsi_device
which successfully return with FCP_RESID_UNDER, FSF_FCP_RSP_AVAILABLE,
and SAM_STAT_GOOD. This floods the HBA trace area and we cannot see
other and important HBA trace records long enough.
Therefore, do not trace HBA response errors for pure benign residual
under counts at the default trace level.
This excludes benign residual under count combined with other validity
bits set in FCP_RSP_IU, such as FCP_SNS_LEN_VAL. For all those other
cases, we still do want to see both the HBA record and the corresponding
SCSI record by default.
Signed-off-by: Steffen Maier <maier@linux.vnet.ibm.com> Fixes: a54ca0f62f95 ("[SCSI] zfcp: Redesign of the debug tracing for HBA records.") Reviewed-by: Benjamin Block <bblock@linux.vnet.ibm.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When SCSI EH invokes zFCP's callbacks for eh_device_reset_handler() and
eh_target_reset_handler(), it expects us to relent the ownership over
the given scsi_cmnd and all other scsi_cmnds within the same scope - LUN
or target - when returning with SUCCESS from the callback ('release'
them). SCSI EH can then reuse those commands.
We did not follow this rule to release commands upon SUCCESS; and if
later a reply arrived for one of those supposed to be released commands,
we would still make use of the scsi_cmnd in our ingress tasklet. This
will at least result in undefined behavior or a kernel panic because of
a wrong kernel pointer dereference.
To fix this, we NULLify all pointers to scsi_cmnds (struct zfcp_fsf_req
*)->data in the matching scope if a TMF was successful. This is done
under the locks (struct zfcp_adapter *)->abort_lock and (struct
zfcp_reqlist *)->lock to prevent the requests from being removed from
the request-hashtable, and the ingress tasklet from making use of the
scsi_cmnd-pointer in zfcp_fsf_fcp_cmnd_handler().
For cases where a reply arrives during SCSI EH, but before we get a
chance to NULLify the pointer - but before we return from the callback
-, we assume that the code is protected from races via the CAS operation
in blk_complete_request() that is called in scsi_done().
The following stacktrace shows an example for a crash resulting from the
previous behavior:
For SRIOV enabled firmware, if there is a OCR(online controller reset)
possibility driver set the convert flag to 1, which is not happening if
there are outstanding commands even after 180 seconds. As driver does
not set convert flag to 1 and still making the OCR to run, VF(Virtual
function) driver is directly writing on to the register instead of
waiting for 30 seconds. Setting convert flag to 1 will cause VF driver
will wait for 30 secs before going for reset.
Signed-off-by: Kiran Kumar Kasturi <kiran-kumar.kasturi@broadcom.com> Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Tomas Henzl <thenzl@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a disagreement between drivers/tty/vt/keyboard.c and
drivers/input/input-leds.c with regard to what is a Scroll Lock LED
trigger name: input calls it "kbd-scrolllock", but vt calls it
"kbd-scrollock" (two l's).
This prevents Scroll Lock LED trigger from binding to this LED by default.
Since it is a scroLL Lock LED, this interface was introduced only about a
year ago and in an Internet search people seem to reference this trigger
only to set it to this LED let's simply rename it to "kbd-scrolllock".
Also, it looks like this was supposed to be changed before this code was
merged: https://lkml.org/lkml/2015/6/9/697 but it was done only on
the input side.
Signed-off-by: Maciej S. Szmigiero <mail@maciej.szmigiero.name> Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a block device is closed while iterate_bdevs() is handling it, the
following NULL pointer dereference occurs because bdev->b_disk is NULL
in bdev_get_queue(), which is called from blk_get_backing_dev_info() (in
turn called by the mapping_cap_writeback_dirty() call in
__filemap_fdatawrite_range()):
The crash is easily reproducible by running the following command, if an
msleep(100) is inserted before the call to func() in iterate_devs():
while :; do head -c1 /dev/nullb0; done > /dev/null & while :; do sync; done
Fix it by holding the bd_mutex across the func() call and only calling
func() if the bdev is opened.
Fixes: 5c0d6b60a0ba ("vfs: Create function for iterating over block devices") Reported-and-tested-by: Wei Fang <fangwei1@huawei.com> Signed-off-by: Rabin Vincent <rabinv@axis.com> Signed-off-by: Jan Kara <jack@suse.cz> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
pm_runtime_autosuspend can take synchronous or asynchronous
paths, Because we are calling pm_runtime_mark_last_busy just before
this most of the cases it takes the asynchronous way. However,
when the FW or driver resets during already running runtime suspend,
the call will result in calling to the driver's rpm callback and results
in a deadlock on device_lock.
The simplest fix is to replace pm_runtime_autosuspend with
asynchronous pm_request_autosuspend.
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com> Signed-off-by: Tomas Winkler <tomas.winkler@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ast_get_dram_info() configures a window in order to access BMC memory.
A BMC register can be configured to disallow this, and if so, causes
an infinite loop in the ast driver which renders the system unusable.
Fix this by erroring out if an error is detected. On powerpc systems with
EEH, this leads to the device being fenced and the system continuing to
operate.
Hook up drm_compat_ioctl to support 32-bit userspace on 64-bit kernels.
It turns out that N2600 and N2800 comes with 64-bit enabled. We
previously assumed there where no such systems out there.
This avoids an issue that occurs when we're attempting to preempt multiple
channels simultaneously. HW seems to ignore preempt requests while it's
still processing a previous one, which, well, makes sense.
Fixes random "fifo: SCHED_ERROR 0d []" + GPCCS page faults during parallel
piglit runs on (at least) GM107.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The HP Pavilion dv6 has a non-working acpi_video0 backlight interface
and an intel_backlight interface which works fine. Add a force_native
quirk for it so that the non-working acpi_video0 interface does not get
registered.
Note that there are quite a few HP Pavilion dv6 variants, some
woth ATI and some with NVIDIA hybrid gfx, both seem to need this
quirk to have working backlight control. There are also some versions
with only Intel integrated gfx, these may not need this quirk, but it
should not hurt there.
The Dell XPS 17 L702X has a non-working acpi_video0 backlight interface
and an intel_backlight interface which works fine. Add a force_native
quirk for it so that the non-working acpi_video0 interface does not get
registered.
Note that there also is an issue with the brightnesskeys on this laptop,
they do not generate key-press events in anyway. That is not solved by
this patch.
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1123661 Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 0557344e2149 ("staging: comedi: ni_mio_common: fix local var for
32-bit read") changed the type of local variable `d` from `unsigned
short` to `unsigned int` to fix a bug introduced in
commit 9c340ac934db ("staging: comedi: ni_stc.h: add read/write
callbacks to struct ni_private") when reading AI data for NI PCI-6110
and PCI-6111 cards. Unfortunately, other parts of the function rely on
the variable being `unsigned short` when an offset value in local
variable `signbits` is added to `d` before writing the value to the
`data` array:
d += signbits;
data[n] = d;
The `signbits` variable will be non-zero in bipolar mode, and is used to
convert the hardware's 2's complement, 16-bit numbers to Comedi's
straight binary sample format (with 0 representing the most negative
voltage). This breaks because `d` is now 32 bits wide instead of 16
bits wide, so after the addition of `signbits`, `data[n]` ends up being
set to values above 65536 for negative voltages. This affects all
supported "E series" cards except PCI-6143 (and PXI-6143). Fix it by
ANDing the value written to the `data[n]` with the mask 0xffff.
Fixes: 0557344e2149 ("staging: comedi: ni_mio_common: fix local var for 32-bit read") Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For NI M Series cards, the Comedi `insn_read` handler for the AI
subdevice is broken due to ANDing the value read from the AI FIFO data
register with an incorrect mask. The incorrect mask clears all but the
most significant bit of the sample data. It should preserve all the
sample data bits. Correct it.
Fixes: 817144ae7fda ("staging: comedi: ni_mio_common: remove unnecessary use of 'board->adbits'") Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the critical sysfs entry the thermal hwmon was returning wrong
temperature to the user-space. It was reporting the temperature of the
first trip point instead of the temperature of critical trip point.
For example:
/sys/class/hwmon/hwmon0/temp1_crit:50000
/sys/class/thermal/thermal_zone0/trip_point_0_temp:50000
/sys/class/thermal/thermal_zone0/trip_point_0_type:active
/sys/class/thermal/thermal_zone0/trip_point_3_temp:120000
/sys/class/thermal/thermal_zone0/trip_point_3_type:critical
Since commit e68b16abd91d ("thermal: add hwmon sysfs I/F") the driver
have been registering a sysfs entry if get_crit_temp() callback was
provided. However when accessed, it was calling get_trip_temp() instead
of the get_crit_temp().
bcm2835_pll_divider_off() is resetting the divider field in the A2W reg
to zero when disabling the clock.
Make sure we preserve this value by reading the previous a2w_reg value
first and ORing the result with A2W_PLL_CHANNEL_DISABLE.
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com> Fixes: 41691b8862e2 ("clk: bcm2835: Add support for programming the audio domain clocks") Reviewed-by: Eric Anholt <eric@anholt.net> Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The clocksource delta to nanoseconds conversion is using signed math, but
the delta is unsigned. This makes the conversion space smaller than
necessary and in case of a multiplication overflow the conversion can
become negative. The conversion is done with scaled math:
Shifting a signed integer right obvioulsy preserves the sign, which has
interesting consequences:
- Time jumps backwards
- __iter_div_u64_rem() which is used in one of the calling code pathes
will take forever to piecewise calculate the seconds/nanoseconds part.
This has been reported by several people with different scenarios:
David observed that when stopping a VM with a debugger:
"It was essentially the stopped by debugger case. I forget exactly why,
but the guest was being explicitly stopped from outside, it wasn't just
scheduling lag. I think it was something in the vicinity of 10 minutes
stopped."
When lifting the stop the machine went dead.
The stopped by debugger case is not really interesting, but nevertheless it
would be a good thing not to die completely.
But this was also observed on a live system by Liav:
"When the OS is too overloaded, delta will get a high enough value for the
msb of the sum delta * tkr->mult + tkr->xtime_nsec to be set, and so
after the shift the nsec variable will gain a value similar to
0xffffffffff000000."
Unfortunately this has been reintroduced recently with commit 6bd58f09e1d8
("time: Add cycles to nanoseconds translation"). It had been fixed a year
ago already in commit 35a4933a8959 ("time: Avoid signed overflow in
timekeeping_get_ns()").
Though it's not surprising that the issue has been reintroduced because the
function itself and the whole call chain uses s64 for the result and the
propagation of it. The change in this recent commit is subtle:
s64 nsec;
- nsec = (d * m + n) >> s:
+ nsec = d * m + n;
+ nsec >>= s;
d being type of cycle_t adds another level of obfuscation.
This wouldn't have happened if the previous change to unsigned computation
would have made the 'nsec' variable u64 right away and a follow up patch
had cleaned up the whole call chain.
There have been patches submitted which basically did a revert of the above
patch leaving everything else unchanged as signed. Back to square one. This
spawned a admittedly pointless discussion about potential users which rely
on the unsigned behaviour until someone pointed out that it had been fixed
before. The changelogs of said patches added further confusion as they made
finally false claims about the consequences for eventual users which expect
signed results.
Despite delta being cycle_t, aka. u64, it's very well possible to hand in
a signed negative value and the signed computation will happily return the
correct result. But nobody actually sat down and analyzed the code which
was added as user after the propably unintended signed conversion.
Though in sensitive code like this it's better to analyze it proper and
make sure that nothing relies on this than hunting the subtle wreckage half
a year later. After analyzing all call chains it stands that no caller can
hand in a negative value (which actually would work due to the s64 cast)
and rely on the signed math to do the right thing.
Change the conversion function to unsigned math. The conversion of all call
chains is done in a follow up patch.
This solves the starvation issue, which was caused by the negative result,
but it does not solve the underlying problem. It merily procrastinates
it. When the timekeeper update is deferred long enough that the unsigned
multiplication overflows, then time going backwards is observable again.
It does neither solve the issue of clocksources with a small counter width
which will wrap around possibly several times and cause random time stamps
to be generated. But those are usually not found on systems used for
virtualization, so this is likely a non issue.
I took the liberty to claim authorship for this simply because
analyzing all callsites and writing the changelog took substantially
more time than just making the simple s/s64/u64/ change and ignore the
rest.
Fixes: 6bd58f09e1d8 ("time: Add cycles to nanoseconds translation") Reported-by: David Gibson <david@gibson.dropbear.id.au> Reported-by: Liav Rehana <liavr@mellanox.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Parit Bhargava <prarit@redhat.com> Cc: Laurent Vivier <lvivier@redhat.com> Cc: "Christopher S. Hall" <christopher.s.hall@intel.com> Cc: Chris Metcalf <cmetcalf@mellanox.com> Cc: Richard Cochran <richardcochran@gmail.com> Cc: John Stultz <john.stultz@linaro.org> Link: http://lkml.kernel.org/r/20161208204228.688545601@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The regulator has never been properly enabled, it has been
dormant all the time. It's strange that MMC was working
at all, but it likely worked by the signals going through
the levelshifter and reaching the card anyways.
Fixes: 3615a34ea1a6 ("regulator: add STw481x VMMC driver") Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Clearing the tuning bits should reset the tuning circuit. However there is
more to do. Reset the command and data lines for good measure, and then
for eMMC ensure the card is not still trying to process a tuning command by
sending a stop command.
Note the JEDEC eMMC specification says the stop command (CMD12) can be used
to stop a tuning command (CMD21) whereas the SD specification is silent on
the subject with respect to the SD tuning command (CMD19). Considering that
CMD12 is not a valid SDIO command, the stop command is sent only when the
tuning command is CMD21 i.e. for eMMC. That addresses cases seen so far
which have been on eMMC.
Note that this replaces the commit fe5fb2e3b58f ("mmc: sdhci: Reset cmd and
data circuits after tuning failure") which is being reverted for v4.9+.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> Tested-by: Dan O'Donovan <dan@emutex.com> Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The active_high LED of my Wistron DNMA-92 is still being recognized as
active_low on 4.7.6 mainline. When I was preparing my former commit 0f9edcdd88a9 ("ath9k: Fix LED polarity for some Mini PCI AR9220 MB92
cards.") to fix that I must have somehow messed up with testing, because
I tested the final version of that patch before sending it, and it was
apparently working; but now it is not working on 4.7.6 mainline.
I initially added the PCI_DEVICE_SUB section for 0x0029/0x2096 above the
PCI_VDEVICE section for 0x0029; but then I moved the former below the
latter after seeing how 0x002A sections were sorted in the file.
This turned out to be wrong: if a generic PCI_VDEVICE entry (that has
both subvendor and subdevice IDs set to PCI_ANY_ID) is put before a more
specific one (PCI_DEVICE_SUB), then the generic PCI_VDEVICE entry will
match first and will be used.
With this patch, 0x0029/0x2096 has finally got active_high LED on 4.7.6.
While I'm at it, let's fix 0x002A too by also moving its generic definition
below its specific ones.
Fixes: 0f9edcdd88a9 ("ath9k: Fix LED polarity for some Mini PCI AR9220 MB92 cards.") Signed-off-by: Vittorio Gambaletta <linuxbugs@vittgam.net>
[kvalo@qca.qualcomm.com: improve the commit log based on email discussions] Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When mac80211 abandons an association attempt, it may free
all the data structures, but inform cfg80211 and userspace
about it only by sending the deauth frame it received, in
which case cfg80211 has no link to the BSS struct that was
used and will not cfg80211_unhold_bss() it.
Fix this by providing a way to inform cfg80211 of this with
the BSS entry passed, so that it can clean up properly, and
use this ability in the appropriate places in mac80211.
This isn't ideal: some code is more or less duplicated and
tracing is missing. However, it's a fairly small change and
it's thus easier to backport - cleanups can come later.
Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In commit a5ffbe0a1993 ("rtlwifi: Fix scheduling while atomic bug") and
commit a269913c52ad ("rtlwifi: Rework rtl_lps_leave() and rtl_lps_enter()
to use work queue"), an error was introduced in the power-save routines
due to the fact that leaving PS was delayed by the use of a work queue.
This problem is fixed by detecting if the enter or leave routines are
in interrupt mode. If so, the workqueue is used to place the request.
If in normal mode, the enter or leave routines are called directly.
Fixes: a269913c52ad ("rtlwifi: Rework rtl_lps_leave() and rtl_lps_enter() to use work queue") Reported-by: Ping-Ke Shih <pkshih@realtek.com> Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When there is a CRC error in the SPROM read from the device, the code
attempts to handle a fallback SPROM. When this also fails, the driver
returns zero rather than an error code.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Matt reported that we have a NULL pointer dereference
in ppp_pernet() from ppp_connect_channel(),
i.e. pch->chan_net is NULL.
This is due to that a parallel ppp_unregister_channel()
could happen while we are in ppp_connect_channel(), during
which pch->chan_net set to NULL. Since we need a reference
to net per channel, it makes sense to sync the refcnt
with the life time of the channel, therefore we should
release this reference when we destroy it.
Fixes: 1f461dcdd296 ("ppp: take reference on channels netns") Reported-by: Matt Bennett <Matt.Bennett@alliedtelesis.co.nz> Cc: Paul Mackerras <paulus@samba.org> Cc: linux-ppp@vger.kernel.org Cc: Guillaume Nault <g.nault@alphalink.fr> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: David S. Miller <davem@davemloft.net> Cc: bmajal222 <bmajal222@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The global mutex of 'gdp_mutex' is used to serialize creating/querying
glue dir and its cleanup. Turns out it isn't a perfect way because
part(kobj_kset_leave()) of the actual cleanup action() is done inside
the release handler of the glue dir kobject. That means gdp_mutex has
to be held before releasing the last reference count of the glue dir
kobject.
This patch moves glue dir's cleanup after kobject_del() in device_del()
for avoiding the race.
Cc: Yijing Wang <wangyijing@huawei.com> Reported-by: Chandra Sekhar Lingutla <clingutla@codeaurora.org> Signed-off-by: Ming Lei <ming.lei@canonical.com> Cc: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
xlog_recover_clear_agi_bucket didn't set the
type to XFS_BLFT_AGI_BUF, so we got a warning during log
replay (or an ASSERT on a debug build).
XFS (md0): Unknown buffer type 0!
XFS (md0): _xfs_buf_ioapply: no ops on block 0xaea8802/0x1
Fix this, as was done in f19b872b for 2 other locations
with the same problem.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function xen_guest_init is using __alloc_percpu with an alignment
which are not power of two.
However, the percpu allocator never supported alignments which are not power
of two and has always behaved incorectly in thise case.
Commit 3ca45a4 "percpu: ensure requested alignment is power of two"
introduced a check which trigger a warning [1] when booting linux-next
on Xen. But in reality this bug was always present.
This can be fixed by replacing the call to __alloc_percpu with
alloc_percpu. The latter will use an alignment which are a power of two.
Commit 9c17d96500f7 ("xen/gntdev: Grant maps should not be subject to
NUMA balancing") set VM_IO flag to prevent grant maps from being
subjected to NUMA balancing.
It was discovered recently that this flag causes get_user_pages() to
always fail with -EFAULT.
We've got a delay loop waiting for secondary CPUs. That loop uses
loops_per_jiffy. However, loops_per_jiffy doesn't actually mean how
many tight loops make up a jiffy on all architectures. It is quite
common to see things like this in the boot log:
Calibrating delay loop (skipped), value calculated using timer
frequency.. 48.00 BogoMIPS (lpj=24000)
In my case I was seeing lots of cases where other CPUs timed out
entering the debugger only to print their stack crawls shortly after the
kdb> prompt was written.
Elsewhere in kgdb we already use udelay(), so that should be safe enough
to use to implement our timeout. We'll delay 1 ms for 1000 times, which
should give us a full second of delay (just like the old code wanted)
but allow us to notice that we're done every 1 ms.
[akpm@linux-foundation.org: simplifications, per Daniel] Link: http://lkml.kernel.org/r/1477091361-2039-1-git-send-email-dianders@chromium.org Signed-off-by: Douglas Anderson <dianders@chromium.org> Reviewed-by: Daniel Thompson <daniel.thompson@linaro.org> Cc: Jason Wessel <jason.wessel@windriver.com> Cc: Brian Norris <briannorris@chromium.org> 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>
NMI handler doesn't call set_irq_regs(), it's set only by normal IRQ.
Thus get_irq_regs() returns NULL or stale registers snapshot with IP/SP
pointing to the code interrupted by IRQ which was interrupted by NMI.
NULL isn't a problem: in this case watchdog calls dump_stack() and
prints full stack trace including NMI. But if we're stuck in IRQ
handler then NMI watchlog will print stack trace without IRQ part at
all.
This patch uses registers snapshot passed into NMI handler as arguments:
these registers point exactly to the instruction interrupted by NMI.
Fixes: 55537871ef66 ("kernel/watchdog.c: perform all-CPU backtrace in case of hard lockup") Link: http://lkml.kernel.org/r/146771764784.86724.6006627197118544150.stgit@buzz Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Ulrich Obergfell <uobergfe@redhat.com> Cc: Aaron Tomlin <atomlin@redhat.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>
We can not unlock/lock cifs_tcp_ses_lock while walking through ses
and tcon lists because it can corrupt list iterator pointers and
a tcon structure can be released if we don't hold an extra reference.
Fix it by moving a reconnect process to a separate delayed work
and acquiring a reference to every tcon that needs to be reconnected.
Also do not send an echo request on newly established connections.
Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In dm_sm_metadata_create() we temporarily change the dm_space_map
operations from 'ops' (whose .destroy function deallocates the
sm_metadata) to 'bootstrap_ops' (whose .destroy function doesn't).
If dm_sm_metadata_create() fails in sm_ll_new_metadata() or
sm_ll_extend(), it exits back to dm_tm_create_internal(), which calls
dm_sm_destroy() with the intention of freeing the sm_metadata, but it
doesn't (because the dm_space_map operations is still set to
'bootstrap_ops').
Fix this by setting the dm_space_map operations back to 'ops' if
dm_sm_metadata_create() fails when it is set to 'bootstrap_ops'.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In crypt_set_key(), if a failure occurs while replacing the old key
(e.g. tfm->setkey() fails) the key must not have DM_CRYPT_KEY_VALID flag
set. Otherwise, the crypto layer would have an invalid key that still
has DM_CRYPT_KEY_VALID flag set.
Signed-off-by: Ondrej Kozina <okozina@redhat.com> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fix to return error code -EINVAL instead of 0, as is done elsewhere in
this function.
Fixes: e80d1c805a3b ("dm: do not override error code returned from dm_get_device()") Signed-off-by: Wei Yongjun <weiyj.lk@gmail.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The meaning of the BLK_MQ_S_STOPPED flag is "do not call
.queue_rq()". Hence modify blk_mq_make_request() such that requests
are queued instead of issued if a queue has been stopped.
Reported-by: Ming Lei <tom.leiming@gmail.com> Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Ming Lei <tom.leiming@gmail.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by: Sagi Grimberg <sagi@grimberg.me> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ep->mult is supposed to be set to Isochronous and
Interrupt Endapoint's multiplier value. This value
is computed from different places depending on the
link speed.
If we're dealing with HighSpeed, then it's part of
bits [12:11] of wMaxPacketSize. This case wasn't
taken into consideration before.
While at that, also make sure the ep->mult defaults
to one so drivers can use it unconditionally and
assume they'll never multiply ep->maxpacket to zero.
Cc: <stable@vger.kernel.org> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When the user namespace support was merged the need to prevent
ptrace from revealing the contents of an unreadable executable
was overlooked.
Correct this oversight by ensuring that the executed file
or files are in mm->user_ns, by adjusting mm->user_ns.
Use the new function privileged_wrt_inode_uidgid to see if
the executable is a member of the user namespace, and as such
if having CAP_SYS_PTRACE in the user namespace should allow
tracing the executable. If not update mm->user_ns to
the parent user namespace until an appropriate parent is found.
Reported-by: Jann Horn <jann@thejh.net> Fixes: 9e4a36ece652 ("userns: Fail exec for suid and sgid binaries with ids outside our user namespace.") Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If you have a process that has set itself to be non-dumpable, and it
then undergoes exec(2), any CLOEXEC file descriptors it has open are
"exposed" during a race window between the dumpable flags of the process
being reset for exec(2) and CLOEXEC being applied to the file
descriptors. This can be exploited by a process by attempting to access
/proc/<pid>/fd/... during this window, without requiring CAP_SYS_PTRACE.
The race in question is after set_dumpable has been (for get_link,
though the trace is basically the same for readlink):
Which will return 0, during the race window and CLOEXEC file descriptors
will still be open during this window because do_close_on_exec has not
been called yet. As a result, the ordering of these calls should be
reversed to avoid this race window.
This is of particular concern to container runtimes, where joining a
PID namespace with file descriptors referring to the host filesystem
can result in security issues (since PRCTL_SET_DUMPABLE doesn't protect
against access of CLOEXEC file descriptors -- file descriptors which may
reference filesystem objects the container shouldn't have access to).
Our system uses significantly more slab memory with memcg enabled with
the latest kernel. With 3.10 kernel, slab uses 2G memory, while with
4.6 kernel, 6G memory is used. The shrinker has problem. Let's see we
have two memcg for one shrinker. In do_shrink_slab:
1. Check cg1. nr_deferred = 0, assume total_scan = 700. batch size
is 1024, then no memory is freed. nr_deferred = 700
2. Check cg2. nr_deferred = 700. Assume freeable = 20, then
total_scan = 10 or 40. Let's assume it's 10. No memory is freed.
nr_deferred = 10.
The deferred share of cg1 is lost in this case. kswapd will free no
memory even run above steps again and again.
The fix makes sure one memcg's deferred share isn't lost.
Link: http://lkml.kernel.org/r/2414be961b5d25892060315fbb56bb19d81d0c07.1476227351.git.shli@fb.com Signed-off-by: Shaohua Li <shli@fb.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Vladimir Davydov <vdavydov@parallels.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>
The struct file_operations instance serving the f2fs/status debugfs file
lacks an initialization of its ->owner.
This means that although that file might have been opened, the f2fs module
can still get removed. Any further operation on that opened file, releasing
included, will cause accesses to unmapped memory.
Indeed, Mike Marshall reported the following:
BUG: unable to handle kernel paging request at ffffffffa0307430
IP: [<ffffffff8132a224>] full_proxy_release+0x24/0x90
<...>
Call Trace:
[] __fput+0xdf/0x1d0
[] ____fput+0xe/0x10
[] task_work_run+0x8e/0xc0
[] do_exit+0x2ae/0xae0
[] ? __audit_syscall_entry+0xae/0x100
[] ? syscall_trace_enter+0x1ca/0x310
[] do_group_exit+0x44/0xc0
[] SyS_exit_group+0x14/0x20
[] do_syscall_64+0x61/0x150
[] entry_SYSCALL64_slow_path+0x25/0x25
<...>
---[ end trace f22ae883fa3ea6b8 ]---
Fixing recursive fault but reboot is needed!
Fix this by initializing the f2fs/status file_operations' ->owner with
THIS_MODULE.
This will allow debugfs to grab a reference to the f2fs module upon any
open on that file, thus preventing it from getting removed.
Fixes: 902829aa0b72 ("f2fs: move proc files to debugfs") Reported-by: Mike Marshall <hubcap@omnibond.com> Reported-by: Martin Brandenburg <martin@omnibond.com> Signed-off-by: Nicolai Stange <nicstange@gmail.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently data journalling is incompatible with encryption: enabling both
at the same time has never been supported by design, and would result in
unpredictable behavior. However, users are not precluded from turning on
both features simultaneously. This change programmatically replaces data
journaling for encrypted regular files with ordered data journaling mode.
Background:
Journaling encrypted data has not been supported because it operates on
buffer heads of the page in the page cache. Namely, when the commit
happens, which could be up to five seconds after caching, the commit
thread uses the buffer heads attached to the page to copy the contents of
the page to the journal. With encryption, it would have been required to
keep the bounce buffer with ciphertext for up to the aforementioned five
seconds, since the page cache can only hold plaintext and could not be
used for journaling. Alternatively, it would be required to setup the
journal to initiate a callback at the commit time to perform deferred
encryption - in this case, not only would the data have to be written
twice, but it would also have to be encrypted twice. This level of
complexity was not justified for a mode that in practice is very rarely
used because of the overhead from the data journalling.
Solution:
If data=journaled has been set as a mount option for a filesystem, or if
journaling is enabled on a regular file, do not perform journaling if the
file is also encrypted, instead fall back to the data=ordered mode for the
file.
Rationale:
The intent is to allow seamless and proper filesystem operation when
journaling and encryption have both been enabled, and have these two
conflicting features gracefully resolved by the filesystem.
Fixes: 67cf5b09a46f ("ext4: add the basic function for inline data support") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The commit "ext4: sanity check the block and cluster size at mount
time" should prevent any problems, but in case the superblock is
modified while the file system is mounted, add an extra safety check
to make sure we won't overrun the allocated buffer.
Fix a large number of problems with how we handle mount options in the
superblock. For one, if the string in the superblock is long enough
that it is not null terminated, we could run off the end of the string
and try to interpret superblocks fields as characters. It's unlikely
this will cause a security problem, but it could result in an invalid
parse. Also, parse_options is destructive to the string, so in some
cases if there is a comma-separated string, it would be modified in
the superblock. (Fortunately it only happens on file systems with a
1k block size.)
The number of 'counters' elements needed in 'struct sg' is
super_block->s_blocksize_bits + 2. Presently we have 16 'counters'
elements in the array. This is insufficient for block sizes >= 32k. In
such cases the memcpy operation performed in ext4_mb_seq_groups_show()
would cause stack memory corruption.
Fixes: c9de560ded61f Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
'border' variable is set to a value of 2 times the block size of the
underlying filesystem. With 64k block size, the resulting value won't
fit into a 16-bit variable. Hence this commit changes the data type of
'border' to 'unsigned int'.
The AEAD givenc descriptor relies on moving the IV through the
output FIFO and then back to the CTX2 for authentication. The
SEQ FIFO STORE could be scheduled before the data can be
read from OFIFO, especially since the SEQ FIFO LOAD needs
to wait for the SEQ FIFO LOAD SKIP to finish first. The
SKIP takes more time when the input is SG than when it's
a contiguous buffer. If the SEQ FIFO LOAD is not scheduled
before the STORE, the DECO will hang waiting for data
to be available in the OFIFO so it can be transferred to C2.
In order to overcome this, first force transfer of IV to C2
by starting the "cryptlen" transfer first and then starting to
store data from OFIFO to the output buffer.
When the flag PT_PTRACE_CAP was added the PTRACE_TRACEME path was
overlooked. This can result in incorrect behavior when an application
like strace traces an exec of a setuid executable.
Further PT_PTRACE_CAP does not have enough information for making good
security decisions as it does not report which user namespace the
capability is in. This has already allowed one mistake through
insufficient granulariy.
I found this issue when I was testing another corner case of exec and
discovered that I could not get strace to set PT_PTRACE_CAP even when
running strace as root with a full set of caps.
This change fixes the above issue with strace allowing stracing as
root a setuid executable without disabling setuid. More fundamentaly
this change allows what is allowable at all times, by using the correct
information in it's decision.
During exec dumpable is cleared if the file that is being executed is
not readable by the user executing the file. A bug in
ptrace_may_access allows reading the file if the executable happens to
enter into a subordinate user namespace (aka clone(CLONE_NEWUSER),
unshare(CLONE_NEWUSER), or setns(fd, CLONE_NEWUSER).
This problem is fixed with only necessary userspace breakage by adding
a user namespace owner to mm_struct, captured at the time of exec, so
it is clear in which user namespace CAP_SYS_PTRACE must be present in
to be able to safely give read permission to the executable.
The function ptrace_may_access is modified to verify that the ptracer
has CAP_SYS_ADMIN in task->mm->user_ns instead of task->cred->user_ns.
This ensures that if the task changes it's cred into a subordinate
user namespace it does not become ptraceable.
The function ptrace_attach is modified to only set PT_PTRACE_CAP when
CAP_SYS_PTRACE is held over task->mm->user_ns. The intent of
PT_PTRACE_CAP is to be a flag to note that whatever permission changes
the task might go through the tracer has sufficient permissions for
it not to be an issue. task->cred->user_ns is always the same
as or descendent of mm->user_ns. Which guarantees that having
CAP_SYS_PTRACE over mm->user_ns is the worst case for the tasks
credentials.
To prevent regressions mm->dumpable and mm->user_ns are not considered
when a task has no mm. As simply failing ptrace_may_attach causes
regressions in privileged applications attempting to read things
such as /proc/<pid>/stat
Acked-by: Kees Cook <keescook@chromium.org> Tested-by: Cyrill Gorcunov <gorcunov@openvz.org> Fixes: 8409cca70561 ("userns: allow ptrace from non-init user namespaces") Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
bdev->bd_contains is not stable before calling __blkdev_get().
When __blkdev_get() is called on a parition with ->bd_openers == 0
it sets
bdev->bd_contains = bdev;
which is not correct for a partition.
After a call to __blkdev_get() succeeds, ->bd_openers will be > 0
and then ->bd_contains is stable.
When FMODE_EXCL is used, blkdev_get() calls
bd_start_claiming() -> bd_prepare_to_claim() -> bd_may_claim()
This call happens before __blkdev_get() is called, so ->bd_contains
is not stable. So bd_may_claim() cannot safely use ->bd_contains.
It currently tries to use it, and this can lead to a BUG_ON().
This happens when a whole device is already open with a bd_holder (in
use by dm in my particular example) and two threads race to open a
partition of that device for the first time, one opening with O_EXCL and
one without.
The thread that doesn't use O_EXCL gets through blkdev_get() to
__blkdev_get(), gains the ->bd_mutex, and sets bdev->bd_contains = bdev;
Immediately thereafter the other thread, using FMODE_EXCL, calls
bd_start_claiming() from blkdev_get(). This should fail because the
whole device has a holder, but because bdev->bd_contains == bdev
bd_may_claim() incorrectly reports success.
This thread continues and blocks on bd_mutex.
The first thread then sets bdev->bd_contains correctly and drops the mutex.
The thread using FMODE_EXCL then continues and when it calls bd_may_claim()
again in:
BUG_ON(!bd_may_claim(bdev, whole, holder));
The BUG_ON fires.
Fix this by removing the dependency on ->bd_contains in
bd_may_claim(). As bd_may_claim() has direct access to the whole
device, it can simply test if the target bdev is the whole device.
Really there's lots of things that can go wrong here, kill all the
BUG_ON()'s and replace the logic ones with ASSERT()'s and return EIO
instead.
Signed-off-by: Josef Bacik <jbacik@fb.com>
[ switched to btrfs_err, errors go to common label ] Reviewed-by: Liu Bo <bo.li.liu@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The extent buffer 'next' needs to be free'd conditionally.
Signed-off-by: Liu Bo <bo.li.liu@oracle.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We don't track the reloc roots in any sort of normal way, so the only way the
root/commit_root nodes get free'd is if the relocation finishes successfully and
the reloc root is deleted. Fix this by free'ing them in free_reloc_roots.
Thanks,
Signed-off-by: Josef Bacik <jbacik@fb.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When relocating tree blocks, we firstly get block information from
back references in the extent tree, we then search fs tree to try to
find all parents of a block.
However, if fs tree is corrupted, eg. if there're some missing
items, we could come across these WARN_ONs and BUG_ONs.
This makes us print some error messages and return gracefully
from balance.
Signed-off-by: Liu Bo <bo.li.liu@oracle.com> Reviewed-by: Josef Bacik <jbacik@fb.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently we allow inconsistence about mixed flag
(BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA).
We'd get ENOSPC if block group has mixed flag and btrfs doesn't.
If that happens, we have one space_info with mixed flag and another
space_info only with BTRFS_BLOCK_GROUP_METADATA, and
global_block_rsv.space_info points to the latter one, but all bytes
from block_group contributes to the mixed space_info, thus all the
allocation will fail with ENOSPC.
This adds a check for the above case.
Reported-by: Vegard Nossum <vegard.nossum@oracle.com> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
[ updated message ] Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>